{
 "model": "claude-sonnet-5-5",
 "judge": "claude-opus-5-5",
 "runs_per_cell": 1,
 "questions": [
  {
   "id": "Q01",
   "type": "single",
   "question": "How many houses were located by the archaeomagnetic prospection at the Neolithic settlement of Vráble, and into how many neighbourhoods are they grouped?",
   "ground_truth": "At least 313 houses, grouped into three spatially distinct neighbourhoods of originally around 15 ha each.",
   "key_facts": [
    "at least 313 houses",
    "three neighbourhoods",
    "about 15 ha each"
   ],
   "approaches": {
    "none": {
     "answer": "I recall that the prospection at Vráble (Slovakia, Linear Pottery culture, c. 5250-4950 BC) located about 313 houses, but I'm not confident of that number. I don't reliably remember how many neighbourhoods they are grouped into, so I won't guess. The published work by Rassmann, Furholt, Müller-Scheeßel and colleagues should give both figures.",
     "verdict": "partial",
     "facts_found": 1,
     "facts_total": 3,
     "judge_note": "Gives roughly 313 houses (tentatively) but explicitly declines to give the number of neighbourhoods and omits the ~15 ha size.",
     "wall_seconds": 11.23,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 689,
     "cache_read_tokens": 0,
     "output_tokens": 728,
     "cost_usd_list": 0.01004,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The archaeomagnetic prospection, carried out by Knut Rassmann in 2010 and 2012, located at least 313 houses at Vráble (Winkelmann et al. 2020). The houses are grouped into three spatially distinct neighbourhoods, each originally about 15 ha in size.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "States at least 313 houses in three neighbourhoods of about 15 ha each; the extra details do not contradict the ground truth.",
     "wall_seconds": 5.7,
     "retrieval_seconds": 2.0,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6852,
     "cache_read_tokens": 0,
     "output_tokens": 99,
     "cost_usd_list": 0.028401999999999997,
     "retrieved_chunks": [
      "W7163143446#2",
      "W7163143446#3",
      "W7163143446#21",
      "W7163143446#34",
      "W7163143446#33",
      "W7163143446#1",
      "W7163143446#0",
      "W7163143446#26"
     ]
    },
    "okf": {
     "answer": "Archaeomagnetic prospection at Vráble located **at least 313 houses**. They are grouped into **three spatially distinct neighbourhoods**, each originally about 15 ha. The settlement covers about 50 ha in total.\n\nOnly 16 of the houses were excavated (2012–2017), so almost 300 remain unexcavated. The source is the bundle's page on the site, sites/vrable.md.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "States at least 313 houses, three neighbourhoods, and about 15 ha each; the extra details do not contradict the ground truth.",
     "wall_seconds": 9.13,
     "retrieval_seconds": null,
     "turns": 3,
     "input_tokens": 6,
     "cache_creation_tokens": 4799,
     "cache_read_tokens": 12042,
     "output_tokens": 379,
     "cost_usd_list": 0.0242022,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "The archaeomagnetic prospection (carried out by Knut Rassmann in 2010 and 2012) identified at least **313 houses** at Vráble-Veľké Lehemby. They are grouped into **three spatially distinct neighbourhoods**, each originally measuring around 15 ha (Furholt et al., *Proceedings of the Prehistoric Society*, 2026).",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer gives at least 313 houses and three neighbourhoods of about 15 ha each; the extra detail does not contradict the ground truth.",
     "wall_seconds": 6.15,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 158808,
     "cache_read_tokens": 0,
     "output_tokens": 128,
     "cost_usd_list": 0.636516,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q02",
   "type": "single",
   "question": "In the Vráble mass deposition, how many headless skeletons had been counted by the end of the 2024 excavations, over what length of the ditch, and what had the 2022 count been?",
   "ground_truth": "At least 77 headless skeletons plus one complete skeleton, all within the first 25 m of the ditch, with the end of the deposition not yet found. In 2022 a minimum of 37 skeletons, 36 of them headless, had been reported.",
   "key_facts": [
    "at least 77 headless skeletons (and one complete skeleton)",
    "within the first 25 m of the ditch",
    "2022: minimum 37 skeletons, 36 headless"
   ],
   "approaches": {
    "none": {
     "answer": "I don't know these figures reliably. I know that Vráble is an Early Neolithic (Linearbandkeramik, c. 5000 BCE) settlement in southwestern Slovakia, studied by a Slovak-German team. I also know that a ditch deposition of skeletons, many apparently lacking skulls, has been reported there. I can't recall the count of headless skeletons at the end of the 2024 excavations, the length of ditch it covered, or the 2022 count. Any numbers I gave would be guesses. The excavation reports, or the publications by the Vráble project team (Nitra Archaeological Institute with the Universities of Kiel and Gothenburg), should give them.",
     "verdict": "abstained",
     "facts_found": 0,
     "facts_total": 3,
     "judge_note": "The candidate declines to give the counts or ditch length, though correct background context is provided.",
     "wall_seconds": 8.89,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 694,
     "cache_read_tokens": 0,
     "output_tokens": 470,
     "cost_usd_list": 0.00748,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "After the 2023 and 2024 excavations, the count stood at **at least 77 headless skeletons plus one complete skeleton**. All were found within the **first 25 m of the ditch** (the outer ditch terminal east of entrance 2). The end of the deposition had not been reached, because the skeletal layer continues eastward.\n\nThe 2022 count was a **minimum of 37 skeletons, 36 of them headless**. The further excavations significantly changed that picture.\n\nThe excerpts attribute the 2022 find to \"entrance 22\" in one place and \"entrance 2\" elsewhere. The latter appears to be the intended reference.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer has all three key facts: at least 77 headless skeletons plus one complete one, the first 25 m of the ditch, and the 2022 count of 37 with 36 headless.",
     "wall_seconds": 3.42,
     "retrieval_seconds": 1.93,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6248,
     "cache_read_tokens": 0,
     "output_tokens": 205,
     "cost_usd_list": 0.027046,
     "retrieved_chunks": [
      "W7163143446#10",
      "W7163143446#5",
      "W7163143446#16",
      "W7163143446#11",
      "W7163143446#18",
      "W7163143446#1",
      "W7163143446#12",
      "W7163143446#13"
     ]
    },
    "okf": {
     "answer": "By the end of the 2024 excavations, at least 77 headless skeletons had been counted, along with one complete skeleton. They lay in the first 25 m of the outer ditch east of entrance 2, and the end of the deposition had not been reached.\n\nThe 2022 count was a minimum of 37 skeletons, of which 36 were headless.\n\nSource: the bundle's finding on the Vráble mass deposition of headless skeletons (`findings/vrable-mass-deposition-headless-skeletons.md`).",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "All three key facts are present: at least 77 headless plus one complete skeleton, the first 25 m of the ditch, and the 2022 minimum of 37 with 36 headless.",
     "wall_seconds": 4.34,
     "retrieval_seconds": null,
     "turns": 2,
     "input_tokens": 4,
     "cache_creation_tokens": 7580,
     "cache_read_tokens": 6140,
     "output_tokens": 262,
     "cost_usd_list": 0.033562,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "By the end of the 2024 excavations (Furholt et al., Proceedings of the Prehistoric Society 2026), the mass deposition in the outer ditch east of entrance 2 comprised **at least 77 headless skeletons plus one complete skeleton**. All of these lay within the **first 25 m of the ditch** from the terminal. The authors had not yet found the end of the deposition: the skeletal layer continues eastward, and a further cluster of at least 12 individuals extended beyond the 2024 excavation area.\n\nThe **2022 count** was a minimum of **37 skeletons, 36 of them headless**, from this section of the ditch. The later campaigns in 2023 and 2024 significantly changed that picture.\n\nWithin the 25 m, 44 of the 77 headless skeletons plus the complete one lay in the first 9.5 m from the terminal.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer gives all three key facts: at least 77 headless skeletons plus one complete skeleton, within the first 25 m of the ditch, and a 2022 minimum of 37 skeletons with 36 headless; the extra details do not contradict the ground truth.",
     "wall_seconds": 4.18,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 594,
     "cache_read_tokens": 158219,
     "output_tokens": 283,
     "cost_usd_list": 0.0210319,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q03",
   "type": "single",
   "question": "How was the Vráble mass deposition dated, and what date range resulted?",
   "ground_truth": "By 14C analysis of rib samples from 12 headless skeletons and of seven single femora; all gave very similar dates of 5300/5210 to 5050/4950 cal BCE.",
   "key_facts": [
    "rib samples from 12 headless skeletons",
    "seven single femora",
    "5300/5210-5050/4950 cal BCE"
   ],
   "approaches": {
    "none": {
     "answer": "I don't know this with enough confidence to answer. Vráble in southwestern Slovakia is best known for its Linearbandkeramik (LBK) settlement and the later Fidvár hillfort. I can't reliably recall how the mass deposition there was dated, or the date range that came out of it. It was probably radiocarbon dating, possibly with Bayesian modelling, but that is an inference, not something I know. I'd rather not give you invented dates. The original excavation publication, or the work of the Vráble team (Nitra Archaeological Institute with Kiel/Frankfurt/Cologne colleagues), would have the exact method and calibrated range.",
     "verdict": "abstained",
     "facts_found": 0,
     "facts_total": 3,
     "judge_note": "The answer declines to commit; its guess of radiocarbon dating is hedged and gives none of the key facts (rib samples, femora, date range).",
     "wall_seconds": 9.7,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 660,
     "cache_read_tokens": 0,
     "output_tokens": 506,
     "cost_usd_list": 0.007704,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The mass deposition was dated by radiocarbon (14C) analysis of rib samples from 12 headless skeletons and from seven single femora. All yielded very similar dates of **5300/5210–5050/4950 cal BCE**. These match the dates from regular burials next to the enclosure ditch and from other headless skeletons outside the mass deposition.\n\nPrecision is limited. The calibration curve is flat between 5210 and 5000 cal BCE, so the deposition could fall anywhere between 5210 cal BCE (about two generations after Vráble was founded) and the end of occupation around 5000–4950 cal BCE.\n\nTaphonomic evidence (little fill sediment under or between bodies, no ditch recuts) suggests the deposition took place over a few years or less, and possibly just hours or days. Finer dating is needed to tell whether this was a recurring practice or a one-off event.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer names 14C dating of ribs from 12 headless skeletons and seven single femora, and gives the exact range 5300/5210–5050/4950 cal BCE.",
     "wall_seconds": 3.94,
     "retrieval_seconds": 1.93,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6660,
     "cache_read_tokens": 0,
     "output_tokens": 292,
     "cost_usd_list": 0.029564,
     "retrieved_chunks": [
      "W7163143446#16",
      "W7163143446#17",
      "W7163143446#1",
      "W7163143446#10",
      "W7204009876#5",
      "W7163143446#18",
      "W7163143446#0",
      "W7163143446#26"
     ]
    },
    "okf": {
     "answer": "The Vráble mass deposition was dated by radiocarbon. The samples were ribs from 12 headless skeletons and seven single femora. All of them gave very similar results, with a range of **5300/5210–5050/4950 cal BCE**. This is consistent with the regular burials near the ditch and with the other headless skeletons.\n\n- **Duration:** There was little fill sediment under or between the superimposed bodies, and no ditch recuts. The authors therefore think deposition probably took \"a few years, maybe even less\", and hours or days are possible.\n- **Position in the settlement's history:** The event could fall anywhere between 5210 cal BCE and the end of occupation at about 5000–4950 cal BCE. 5210 cal BCE is two generations after the founding of Vráble.\n- **Limit on precision:** The calibration curve has a plateau between 5210 and 5000 cal BCE. This prevents a finer chronology. It also leaves open whether this was a one-off event or a recurring long-term practice.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer gives radiocarbon dating of ribs from 12 headless skeletons and seven single femora, with the date range 5300/5210–5050/4950 cal BCE.",
     "wall_seconds": 7.0,
     "retrieval_seconds": null,
     "turns": 3,
     "input_tokens": 6,
     "cache_creation_tokens": 4666,
     "cache_read_tokens": 12586,
     "output_tokens": 592,
     "cost_usd_list": 0.025854600000000002,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "**Method:** The mass deposition at Vráble-Veľké Lehemby was dated by radiocarbon (14C) analysis of rib samples from 12 headless skeletons and of seven single femora (Furholt et al. 2026).\n\n**Result:** All samples gave very similar dates, c. 5300/5210–5050/4950 cal BCE. These agree with the dates from the regular burials next to the enclosure ditch and from the other headless skeletons outside the mass deposition.\n\n**Limits on precision:** The deposition could have happened any time between 5210 cal BCE (about two generations after Vráble's founding) and the end of the site's occupation around 5000–4950 cal BCE. A flat section of the calibration curve between 5210 and 5000 cal BCE prevents a tighter chronology. The authors still judge that the deposition did not take place over a long period, mainly because little fill sediment accumulated under or between the bodies and no ditch recuts are visible. They suggest a few years, perhaps less, and note that hours or days are possible. They also say that deciding whether this was a one-off event or a recurring practice needs more precise dating.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer names 14C dating of rib samples from 12 headless skeletons and seven single femora, giving 5300/5210–5050/4950 cal BCE, which matches the ground truth.",
     "wall_seconds": 4.55,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 560,
     "cache_read_tokens": 158219,
     "output_tokens": 373,
     "cost_usd_list": 0.0217959,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q04",
   "type": "single",
   "question": "How many soil samples were analysed from the two LBK sites in southeastern Poland (Łańcut and Zwięczyca), and how were the plant remains extracted?",
   "ground_truth": "155 soil samples from the LBK phases: 97 from Łańcut (1,142 L) and 57 from Zwięczyca (679 L), plus one sample of mixed LBK and Malice pottery. The plant material was extracted by flotation using 0.5 and 1 mm mesh sieves.",
   "key_facts": [
    "155 samples",
    "97 from Łańcut, 57 from Zwięczyca",
    "flotation",
    "0.5 mm and 1 mm mesh"
   ],
   "approaches": {
    "none": {
     "answer": "I don't know this one. I can't reliably recall the specific number of soil samples analysed from the Łańcut and Zwięczyca LBK (Linear Pottery culture) sites in southeastern Poland, or the extraction method used for the plant remains. The question seems to refer to a particular archaeobotanical paper, and I don't want to guess at figures or at whether flotation, water sieving, or some other technique was used.\n\nIf you can share the paper's title or authors, or paste the relevant methods section, I can help interpret or summarise it.",
     "verdict": "abstained",
     "facts_found": 0,
     "facts_total": 4,
     "judge_note": "The candidate declines to answer and gives no sample counts or extraction method, although the question is answerable.",
     "wall_seconds": 6.35,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 690,
     "cache_read_tokens": 0,
     "output_tokens": 173,
     "cost_usd_list": 0.004494,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "**Number of samples:** 155 soil samples from LBK phases were analysed: 97 from Łańcut (1,142 L of sediment) and 57 from Zwięczyca (679 L). The excerpt also mentions one additional sample (feature no. 102) containing LBK pottery mixed with Malice culture material, which was analysed separately.\n\n**Extraction method:** Plant material was recovered by flotation, using 0.5 mm and 1 mm mesh sieves, at the W. Szafer Institute of Botany, Polish Academy of Sciences (IB PAS). The material was then examined under a binocular microscope at 10× and 30× magnification. Charred remains were identified with the morphological-comparative method, using keys, atlases and reference collections. Only charred material was treated as contemporaneous with the features.",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "The answer gives 155 samples (97 Łańcut, 57 Zwięczyca) and flotation with 0.5 and 1 mm sieves, and adds accurate extra detail.",
     "wall_seconds": 3.55,
     "retrieval_seconds": 1.78,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 7709,
     "cache_read_tokens": 0,
     "output_tokens": 268,
     "cost_usd_list": 0.03352,
     "retrieved_chunks": [
      "W7166088190#4",
      "W7166088190#0",
      "W7166088190#2",
      "W7166088190#3",
      "W7166088190#5",
      "W7166088190#13",
      "W7166088190#9",
      "W7166088190#6"
     ]
    },
    "okf": {
     "answer": "**155 LBK soil samples** were analysed: 97 from Łańcut (1,142 L of soil) and 57 from Zwięczyca (679 L). There was also one mixed LBK/Malice sample, from feature 102. The samples came from long pits associated with houses and from other settlement pits.\n\nThe plant remains were extracted by **flotation**, using 0.5 mm and 1 mm sieves. This was done at the W. Szafer Institute of Botany, Polish Academy of Sciences. The remains were then sorted and identified under a binocular microscope at 10× and 30×. Only charred material was treated as contemporaneous with the features.",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "The answer gives 155 samples (97 from Łańcut, 57 from Zwięczyca) and flotation with 0.5 mm and 1 mm sieves, matching the ground truth; the extra details do not contradict it.",
     "wall_seconds": 6.5,
     "retrieval_seconds": null,
     "turns": 3,
     "input_tokens": 6,
     "cache_creation_tokens": 4158,
     "cache_read_tokens": 12298,
     "output_tokens": 459,
     "cost_usd_list": 0.0224638,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "**Samples:** A total of 155 soil samples from LBK phases were analysed: 97 from Łańcut (1,142 L of sediment) and 57 from Zwięczyca (679 L). One additional sample from Łańcut (feature 102), which contained LBK pottery mixed with Malice culture material, was also analysed.\n\n**Extraction:** Plant material was recovered by flotation using 0.5 mm and 1 mm mesh sieves at the W. Szafer Institute of Botany, Polish Academy of Sciences (IB PAS). The material was then examined under a binocular microscope at 10× and 30× magnification. Charred remains were identified with the morphological-comparative method, using keys and atlases (Kulpa 1974; Bojňanský and Fargašová 2007; Cappers et al. 2009) and the IB PAS reference collection. Only charred material was treated as contemporaneous with the features.",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "The answer has all key facts: 155 samples (97 Łańcut, 57 Zwięczyca), extraction by flotation, and 0.5 and 1 mm mesh; the extra details don't contradict the ground truth.",
     "wall_seconds": 3.91,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 590,
     "cache_read_tokens": 158219,
     "output_tokens": 303,
     "cost_usd_list": 0.0212159,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q05",
   "type": "single",
   "question": "Were hazelnut remains found at the Polish LBK sites, and if so, how many and where?",
   "ground_truth": "More than 100 fragments of hazelnut (Corylus avellana) shell were found at Łańcut; hazelnut was absent at Zwięczyca. The authors suggest the nuts were gathered, because of their high fat and protein content.",
   "key_facts": [
    "more than 100 fragments",
    "at Łańcut",
    "absent at Zwięczyca"
   ],
   "approaches": {
    "none": {
     "answer": "I can't give a reliable answer. Hazelnut shell fragments are a common find on LBK sites across central Europe, so it's likely some turned up at Polish sites such as those in Lesser Poland (the Kraków-Nowa Huta area), Silesia, or Kuyavia. But I don't remember specific counts or site names from the Polish archaeobotanical literature, and I'd be guessing if I gave numbers.\n\nWork by Polish archaeobotanists such as Lityńska-Zając and Wasylikowa, and regional syntheses of Neolithic plant use in Poland, would likely have the site-by-site figures. If you tell me which publication you're working from, I can help interpret it.",
     "verdict": "abstained",
     "facts_found": 0,
     "facts_total": 3,
     "judge_note": "The answer declines to give counts or sites and mentions none of Łańcut, the >100 fragments, or the absence at Zwięczyca.",
     "wall_seconds": 10.07,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 667,
     "cache_read_tokens": 0,
     "output_tokens": 521,
     "cost_usd_list": 0.007881999999999998,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "Yes, but only at one of the two new sites. The excerpts report a high number of charred hazelnut (Corylus avellana) shell fragments at Łańcut site 3, while hazelnuts were absent at Rzeszów-Zwięczyca site 3 (Fig. 4). The excerpts do not give an exact count of fragments.\n\nThis is unusual for LBK sites in Poland. There, hazel is far more often recorded as charcoal than as nutshell, and carpological assemblages usually contain only single fragments (e.g., Lityńska-Zając et al. 2017; Kapcia et al. 2024). The authors suggest the Łańcut finds may indicate deliberate gathering of hazelnuts for their high fat and protein content. The abstract likewise describes a \"relatively high share\" of hazelnut remains in LBK contexts, suggesting the wild species was likely eaten by early farmers.\n\nThe excerpts do not provide a precise number, so the exact quantity would need to be checked in the paper's Fig. 4 or its tables.",
     "verdict": "partial",
     "facts_found": 2,
     "facts_total": 3,
     "judge_note": "The answer correctly places the hazelnut finds at Łańcut and their absence at Zwięczyca, but gives only 'a high number' and not 'more than 100 fragments'.",
     "wall_seconds": 4.25,
     "retrieval_seconds": 1.72,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 7226,
     "cache_read_tokens": 0,
     "output_tokens": 343,
     "cost_usd_list": 0.032338,
     "retrieved_chunks": [
      "W7166088190#15",
      "W7166088190#12",
      "W7166088190#0",
      "W7166088190#4",
      "W7166088190#11",
      "W7166088190#1",
      "W7166088190#2",
      "W7163143446#13"
     ]
    },
    "okf": {
     "answer": "Yes. Hazelnut (*Corylus avellana*) remains were found at Łańcut site 3, an LBK settlement on the Mikośka River in southeastern Poland. There were more than 100 charred shell fragments there, which suggests the LBK farmers gathered them. No hazelnut was found at the other site, Rzeszów-Zwięczyca site 3. The bundle gives only the approximate figure \">100\", not an exact count.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer states more than 100 shell fragments at Łańcut and none at Zwięczyca, matching the ground truth.",
     "wall_seconds": 4.55,
     "retrieval_seconds": null,
     "turns": 2,
     "input_tokens": 4,
     "cache_creation_tokens": 1788,
     "cache_read_tokens": 6113,
     "output_tokens": 270,
     "cost_usd_list": 0.0104713,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "Yes, but only at one of the two sites. Juźwińska et al. report more than 100 fragments of charred hazelnut (*Corylus avellana*) shell from Łańcut site 3. None were found at Rzeszów-Zwięczyca site 3.\n\nThe authors note that hazel at most Polish LBK sites occurs mainly as charcoal. Nutshells usually appear only as single fragments. The relatively high number at Łańcut may indicate that hazelnuts were deliberately gathered, probably for their high fat and protein content.\n\nThe paper gives only an approximate count (\"more than 100 fragments\"), not an exact figure.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer states more than 100 hazelnut shell fragments at Łańcut and none at Zwięczyca, and it matches the gathering interpretation.",
     "wall_seconds": 3.61,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 567,
     "cache_read_tokens": 158219,
     "output_tokens": 209,
     "cost_usd_list": 0.020183899999999998,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q06",
   "type": "single",
   "question": "What gap between the end of the LBK and the arrival of the Early Trypillians did the Bayesian interval query estimate for the Southern Buh and Dniester region?",
   "ground_truth": "The interval function gave a gap of 371 to 602 years (95.4% probability), or 418 to 524 years at 68.3%.",
   "key_facts": [
    "371-602 years at 95.4%",
    "418-524 years at 68.3%"
   ],
   "approaches": {
    "none": {
     "answer": "I don't know the figure. The question seems to refer to a specific radiocarbon study that used Bayesian interval queries on the Southern Buh-Dniester region. I can't recall the gap it estimated between the end of the LBK (Linear Pottery culture) and the arrival of Early Trypillian groups, and I don't want to guess a number. If you can share the paper or its title, I can help interpret it.",
     "verdict": "abstained",
     "facts_found": 0,
     "facts_total": 2,
     "judge_note": "The candidate declined to give a figure even though the question is answerable from the source.",
     "wall_seconds": 7.53,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 690,
     "cache_read_tokens": 0,
     "output_tokens": 277,
     "cost_usd_list": 0.005534,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The Bayesian Interval query (Kiosak & Kotova, W7204009876) estimated a pronounced gap between the end of the LBK and the start of Early Trypillia of **371–602 years at 95.4% probability**, and more narrowly **418–524 years at 68.3% probability**.\n\nCeramic hunter-gatherer dates fill this gap, with partial overlap into both the late LBK and early Trypillian phases. The authors' general description is a gap of roughly 300–400 years. In the simplified sequential model, the LBK phase spans 5213–5050 cal BC (95.4%) and the Early Trypillian phase starts at 4676–4662 cal BC (95.4%).",
     "verdict": "correct",
     "facts_found": 2,
     "facts_total": 2,
     "judge_note": "Both intervals match the ground truth: 371–602 years at 95.4% and 418–524 years at 68.3%. The extra contextual details do not contradict it.",
     "wall_seconds": 3.56,
     "retrieval_seconds": 2.13,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 7041,
     "cache_read_tokens": 0,
     "output_tokens": 230,
     "cost_usd_list": 0.030468000000000002,
     "retrieved_chunks": [
      "W7204009876#3",
      "W7204009876#14",
      "W7204009876#16",
      "W7225183261#33",
      "W7204009876#13",
      "W7204009876#0",
      "W7225183261#30",
      "W7225183261#32"
     ]
    },
    "okf": {
     "answer": "The OxCal Interval query put the gap between the end of the LBK and the arrival of the Early Trypillians at **371–602 years** at 95.4% probability. At 68.3% it gave **418–524 years**. These figures come from the overlapping-phase model.\n\nThe paper's abstract and conclusion round this to roughly 300–400 years. That is lower than the model's own intervals.\n\nHunter-gatherer dates fill the interval, which the paper describes as fitting \"like a well-fitting cork\". This is from Kiosak & Kotova 2026.",
     "verdict": "correct",
     "facts_found": 2,
     "facts_total": 2,
     "judge_note": "Both intervals (371-602 years at 95.4% and 418-524 years at 68.3%) are stated correctly, and the extra details do not contradict the ground truth.",
     "wall_seconds": 6.38,
     "retrieval_seconds": null,
     "turns": 3,
     "input_tokens": 6,
     "cache_creation_tokens": 5346,
     "cache_read_tokens": 13154,
     "output_tokens": 438,
     "cost_usd_list": 0.0270914,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "Kiosak and Kotova (paper on the post-LBK hunter-gatherer \"resurgence\") applied OxCal's Interval query to the LBK, ceramic hunter-gatherer, and Early Trypillian phases. It estimated a gap between the LBK and Early Trypillia of:\n\n- **371–602 years** at 95.4% probability\n- **418–524 years** at 68.3% probability\n\nThe authors' abstract and conclusion describe this more loosely as a gap of roughly 300–400 years. Dates from ceramic hunter-gatherer sites (many with Savran-style pottery) fall into this interval.\n\nWhen the three cultural phases were instead modelled as strictly sequential, the LBK phase spans 5213–5050 cal BC (95.4%) and Early Trypillia starts at 4676–4662 cal BC (95.4%), with hunter-gatherers filling the interval between.",
     "verdict": "correct",
     "facts_found": 2,
     "facts_total": 2,
     "judge_note": "Both intervals (371–602 years at 95.4% and 418–524 years at 68.3%) are stated correctly; the extra detail does not contradict the ground truth.",
     "wall_seconds": 4.2,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 590,
     "cache_read_tokens": 158219,
     "output_tokens": 297,
     "cost_usd_list": 0.021155900000000002,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q07",
   "type": "single",
   "question": "How many LBK inhumations did the Alsace Plain dataset contain, from how many sites, and how many Grossgartach inhumations?",
   "ground_truth": "LBK: 297 inhumations (minimum number of individuals 305) from 30 sites. Grossgartach: 228 inhumations (MNI 236) from nine sites.",
   "key_facts": [
    "297 LBK inhumations (MNI 305)",
    "30 sites",
    "228 Grossgartach inhumations",
    "nine sites"
   ],
   "approaches": {
    "none": {
     "answer": "I don't know these figures reliably. I can't recall the exact counts of LBK and Grossgartach inhumations, or the number of LBK sites, in the Alsace Plain corpus. You may be thinking of a particular study, such as a paleodemographic, isotopic or mortuary-practice analysis of Alsatian Early and Middle Neolithic burials. Different studies define the corpus differently, so any number I gave would be a guess. If you tell me the paper or author, or paste the relevant passage, I can help interpret it.",
     "verdict": "abstained",
     "facts_found": 0,
     "facts_total": 4,
     "judge_note": "The answer declines to give any figures even though the question is answerable from the ground truth.",
     "wall_seconds": 8.44,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 681,
     "cache_read_tokens": 0,
     "output_tokens": 360,
     "cost_usd_list": 0.006328,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "According to the excerpt (Waldvogel, W7202185506), the Alsace Plain corpus contained:\n\n- **Linearbandkeramik (LBK):** 297 inhumations (MNI: 305) from 30 sites.\n- **Grossgartach:** 228 inhumations (MNI: 236) from nine sites.\n\nFor comparison, the corpus also included two Hinkelstein inhumations from two sites and 22 Rössen inhumations (MNI: 26) from five sites.",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "All key figures (297/305 LBK from 30 sites, 228/236 Grossgartach from nine sites) match the ground truth.",
     "wall_seconds": 3.11,
     "retrieval_seconds": 1.96,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 7021,
     "cache_read_tokens": 0,
     "output_tokens": 162,
     "cost_usd_list": 0.029708000000000002,
     "retrieved_chunks": [
      "W7202185506#17",
      "W7202185506#16",
      "W7202185506#18",
      "W7202185506#9",
      "W7202185506#19",
      "W7202185506#27",
      "W7202185506#23",
      "W7202185506#30"
     ]
    },
    "okf": {
     "answer": "The Alsace Plain corpus (Waldvogel 2026) contained **297 LBK inhumations** (minimum number of individuals, MNI, 305) from **30 sites**. It also contained **228 Grossgartach inhumations** (MNI 236) from 9 sites.",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "All counts match: 297 LBK (MNI 305) from 30 sites, 228 Grossgartach from nine sites.",
     "wall_seconds": 5.94,
     "retrieval_seconds": null,
     "turns": 3,
     "input_tokens": 6,
     "cache_creation_tokens": 4039,
     "cache_read_tokens": 11568,
     "output_tokens": 386,
     "cost_usd_list": 0.0211848,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "According to Waldvogel's study of the Alsace Plain (Proceedings of the Prehistoric Society, 2026):\n\n- **Linearbandkeramik (LBK):** 297 inhumations (Minimum Number of Individuals: 305) from 30 sites.\n- **Grossgartach:** 228 inhumations (MNI: 236) from nine sites.\n\nFor context, the corpus also included two Hinkelstein inhumations (two sites) and 22 Rössen inhumations (MNI 26, five sites). Cremations were excluded, as were undated remains outside funerary contexts.",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "All key figures (297 LBK inhumations with MNI 305 from 30 sites; 228 Grossgartach inhumations from nine sites) match the ground truth, and the extra detail does not contradict it.",
     "wall_seconds": 3.38,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 581,
     "cache_read_tokens": 158219,
     "output_tokens": 196,
     "cost_usd_list": 0.0201099,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q08",
   "type": "single",
   "question": "How does body orientation in LBK graves differ between Lower and Upper Alsace?",
   "ground_truth": "In Lower Alsace 31% of individuals with known orientation have the head to the west ('antipodic'), contrary to the usual LBK head-east practice. In Upper Alsace 92% have the head to the east, as in the Danube Basin.",
   "key_facts": [
    "Lower Alsace 31% head west",
    "Upper Alsace 92% head east"
   ],
   "approaches": {
    "none": {
     "answer": "I don't know this reliably enough to give specifics. I recall that LBK burials in Alsace show regional differences in body orientation, with Lower Alsace (Bas-Rhin, e.g. Vendenheim, Lingolsheim) and Upper Alsace (Haut-Rhin, e.g. Rixheim, Ensisheim) following different conventions. I can't say which compass directions go with which region, and a guess could be wrong. Christian Jeunesse's work on Early Neolithic funerary practices in Alsace, and the publications on the Alsatian LBK cemeteries, should cover this directly.",
     "verdict": "abstained",
     "facts_found": 0,
     "facts_total": 2,
     "judge_note": "The answer mentions regional differences but explicitly declines to give the directions or percentages.",
     "wall_seconds": 10.38,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 664,
     "cache_read_tokens": 0,
     "output_tokens": 673,
     "cost_usd_list": 0.009389999999999999,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "Body orientation differs sharply between the two regions (Waldvogel, W7202185506):\n\n- **Lower Alsace:** 31% of individuals with known orientation were buried with their heads to the west (\"antipodic\"). This runs contrary to the typical LBK practice of placing the head to the east.\n- **Upper Alsace:** 92% of individuals with known orientation had their heads to the east. This matches the practice documented in the Danube Basin.\n\nThe divergence is clearly visible from phase IV onwards, when the funerary record is relatively balanced between the two regions. Within each region the pattern holds across cemeteries, settlements and undetermined contexts, and it does not appear linked to the age or biological sex of the deceased. Elsewhere, the paper describes Facies I (Upper Alsace and the Paris Basin) as having heads predominantly between NE and SE.",
     "verdict": "correct",
     "facts_found": 2,
     "facts_total": 2,
     "judge_note": "Both key facts are present: Lower Alsace has 31% of burials with the head to the west, and Upper Alsace has 92% with the head to the east, with no contradictions.",
     "wall_seconds": 3.88,
     "retrieval_seconds": 1.9,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6845,
     "cache_read_tokens": 0,
     "output_tokens": 284,
     "cost_usd_list": 0.030224,
     "retrieved_chunks": [
      "W7202185506#10",
      "W7202185506#2",
      "W7202185506#16",
      "W7202185506#9",
      "W7202185506#29",
      "W7202185506#27",
      "W7202185506#24",
      "W7202185506#23"
     ]
    },
    "okf": {
     "answer": "In LBK graves, Lower Alsace and Upper Alsace differ clearly in body orientation:\n\n- **Lower Alsace:** 31% of individuals with known orientation have the head to the west. This is called \"antipodic\" orientation, and it runs against the usual LBK norm of head to the east. Another figure in the bundle gives 31.5% for Facies II.\n- **Upper Alsace:** 92% of individuals with known orientation have the head to the east. This matches the Danube Basin norm. In the Facies II definition (heads between SSW and WNW), antipodic burials are only about 7%.\n\nSchwetzingen shows the same 31% antipodic share as Lower Alsace.\n\nSources: findings/lbk-body-orientation-alsace.md and terms/antipodic-orientation.md.",
     "verdict": "correct",
     "facts_found": 2,
     "facts_total": 2,
     "judge_note": "Both key facts are present (Lower Alsace 31% head west/antipodic, Upper Alsace 92% head east); the extra details do not contradict the ground truth.",
     "wall_seconds": 5.28,
     "retrieval_seconds": null,
     "turns": 2,
     "input_tokens": 4,
     "cache_creation_tokens": 2854,
     "cache_read_tokens": 6110,
     "output_tokens": 385,
     "cost_usd_list": 0.015885,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "According to Waldvogel (2026), the two regions differ clearly in head orientation:\n\n- **Lower Alsace:** 31% of individuals with known orientation have their heads to the west (\"antipodic\"), contrary to the usual LBK practice of heads to the east. The paper's regional comparison puts this at 31.5% in Lower Alsace, and notes a similar 31% at Schwetzingen in the Neckar region.\n- **Upper Alsace:** 92% of individuals with known orientation are buried with heads to the east, in line with practice in the Danube Basin. (Elsewhere the paper phrases this as only 7% being antipodic.)\n\nThe difference is visible from phase IV onwards, when the funerary record is relatively balanced between the two regions. The paper treats it as part of a broader Upper/Lower Alsace distinction, which also appears in other domains, such as house plans. It places Upper Alsace in funerary Facies I and Lower Alsace in Facies II. It argues the patterns do not correlate with sex or age at death.",
     "verdict": "correct",
     "facts_found": 2,
     "facts_total": 2,
     "judge_note": "The answer gives 31% head-west for Lower Alsace and 92% head-east for Upper Alsace, matching both key facts; the extra details do not contradict the ground truth.",
     "wall_seconds": 4.69,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 564,
     "cache_read_tokens": 158219,
     "output_tokens": 341,
     "cost_usd_list": 0.0214919,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q09",
   "type": "cross",
   "question": "How do the papers interpret the function of LBK enclosures in the Rhineland, in Hesbaye and at Vráble?",
   "ground_truth": "Rhineland (Kroon et al.): enclosures appear from house generation XII onward; their size, slope location, multiple entrances and unclosed ring ditches make fortifications unlikely, a socio-ritual gathering place is considered, and their late appearance may signal stress or violence. Hesbaye: likely defensive, with the defensive value coming from the palisade rather than the ditches. Vráble (Furholt et al.): a 1.3 km double ditch with at least six entrances surrounds only one of the three neighbourhoods and the entrances face away from the others, so the enclosure is seen as directed against other members of the same community rather than an outside threat; human bodies were deposited in the ditches.",
   "key_facts": [
    "Rhineland: unlikely fortifications, possibly socio-ritual",
    "Hesbaye: likely defensive, palisade",
    "Vráble: encloses one of three neighbourhoods, entrances face away, directed against own community"
   ],
   "approaches": {
    "none": {
     "answer": "I can't see the papers you mean, so this is the general scholarly picture from memory and may not match their arguments.\n\n- **Rhineland (Aldenhoven Plateau, Merzbach valley, Köln-Lindenthal):** Late LBK enclosures (c. 5200-5000 BC) were long read as defensive, because of the ditches, palisades and violence evidence in the wider LBK. More recent work stresses ritual, communal or ceremonial uses, such as gathering places and boundary-marking, with defence only part of the picture.\n\n- **Hesbaye (Belgium; Darion, Remicourt, Waremme):** Interrupted ditches, often built as discontinuous segments, are usually seen as symbolic or social boundaries around settlement areas rather than fortifications. Some scholars add a role in regional identity or in demarcating a community during the Late/Final LBK.\n\n- **Vráble (Slovakia, c. 5250-4950 BC):** Furholt, Müller, Hofmann and colleagues see the ditches as a way of structuring a large settlement. They treat them as social boundaries between neighbourhoods or groups of houses, not primarily defence.\n\nI can't give exact page-level claims without the texts.",
     "verdict": "partial",
     "facts_found": 1,
     "facts_total": 3,
     "judge_note": "The Rhineland reading (ritual/gathering, defence downplayed) roughly matches, but Hesbaye is given as symbolic boundaries rather than defensive with a palisade, which contradicts the ground truth, and Vráble omits that the enclosure surrounds one of three neighbourhoods with entrances facing away.",
     "wall_seconds": 11.9,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 677,
     "cache_read_tokens": 0,
     "output_tokens": 1009,
     "cost_usd_list": 0.012802,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "**Rhineland:** The function is debated. Enclosures appear only late, from HG XII onward. Early studies read them as fortifications (Buttler & Haberey 1936; Höckmann 1975). Peters (2024) considers that unlikely given their size, slope location, multiple entrances and unclosed ring ditches. They are commonly seen as communal monuments built through group cooperation (Kerig 2003), and a socio-ritual gathering-place interpretation has to be considered. Their late emergence may also signal societal stress or violence (Kondor et al. 2024).\n\n**Hesbaye:** The enclosures are likely defensive, though alternatives can't be excluded. Their defensive value comes from the associated palisade, not from the discontinuous ditches or simple earthworks. The ditches probably resulted from building the palisade, much like the loam-extraction pits beside houses, rather than being a deliberate moat. The complex entrance features at Darion and Waremme-Longchamps also point to defence, though they may have had multiple functions.\n\n**Vráble:** The excerpts do not mention Vráble, so I can't say how they interpret enclosures there.",
     "verdict": "partial",
     "facts_found": 2,
     "facts_total": 3,
     "judge_note": "Rhineland and Hesbaye interpretations are correct, but the answer says nothing about Vráble beyond claiming the excerpts do not cover it.",
     "wall_seconds": 6.07,
     "retrieval_seconds": 2.01,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6757,
     "cache_read_tokens": 0,
     "output_tokens": 659,
     "cost_usd_list": 0.033622,
     "retrieved_chunks": [
      "W7225183261#28",
      "W7225183261#10",
      "W7225183261#29",
      "W7225183261#6",
      "W7225183261#11",
      "W7225183261#7",
      "W7225183261#30",
      "W7225183261#3"
     ]
    },
    "okf": {
     "answer": "The papers don't agree on one function, and the interpretations differ by region.\n\n- **Rhineland:** The enclosures are read as communal monuments and socio-ritual gathering places. The papers cite Kerig 2003, Whittle 1996, Kaufmann 1997 and Haack 2014 on the significance of digging and backfilling the ditches. Older readings as fortifications (Buttler & Haberey 1936; Höckmann 1975) are judged unlikely. Peters 2024 gives the reasons: the enclosures are large, sit on slopes, have multiple entrances and have unclosed ring ditches. Their late appearance, from HG XII, may signal stress or violence.\n- **Hesbaye:** The enclosures are likely defensive. The palisade is the defensive element, and the ditches may be a byproduct of building it (Broes & Bosquet 2007).\n- **Vráble:** The 1.3 km double ditch has at least six entrances and encloses only the south-western neighbourhood, one of three. The entrances face away from the other neighbourhoods. Furholt et al. 2026 therefore see the enclosure as probably not a defence against outside threat. They suggest it responds to tension within the community. Headless skeletons were deposited along at least 320 m of the southern ditch.\n\nKroon et al. 2026 (Rhineland and Hesbaye) expect the debate to continue.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer covers all three regions accurately: Rhineland enclosures as unlikely fortifications and socio-ritual places, Hesbaye as defensive via the palisade, and Vráble as enclosing one of three neighbourhoods with entrances facing away, directed at tension within the community.",
     "wall_seconds": 13.19,
     "retrieval_seconds": null,
     "turns": 8,
     "input_tokens": 8,
     "cache_creation_tokens": 5288,
     "cache_read_tokens": 17578,
     "output_tokens": 1470,
     "cost_usd_list": 0.0376258,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "The papers give different readings for each area.\n\n**Rhineland (Kroon et al.):** Enclosures appear only from house generation HG XII onward. They are seen as communal monuments built through cooperation among larger groups (Kerig 2003). Early views of them as fortifications are considered unlikely, given their size, slope location, multiple entrances and unclosed ring ditches. Socio-ritual gathering places are favoured, though the late appearance may signal societal stress or violence. The authors say the debate will likely continue.\n\n**Hesbaye (Kroon et al.):** Enclosures are likely defensive, though other interpretations are not excluded. The defensive value is thought to come from the palisade rather than the discontinuous ditches. The ditches probably result from palisade construction rather than a deliberate moat. Complex entrance features at Darion and Waremme-Longchamps support this, though multiple functions are possible.\n\n**Vráble (Furholt et al.):** The enclosure surrounds only one of three contemporaneous neighbourhoods, and its entrances face away from the others. The authors therefore argue it is not primarily directed against an outside threat, but against other members of the same settlement community. They link it to intra-community tensions over settlement delineation and group membership. They also suggest it was a space for ritual or magical manipulation of community space, with headless bodies deposited in the ditch.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "All three key facts are present: Rhineland unlikely fortifications with socio-ritual use favoured, Hesbaye defensive via the palisade, and Vráble enclosing one of three neighbourhoods with entrances facing away against its own community.",
     "wall_seconds": 5.24,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 577,
     "cache_read_tokens": 158219,
     "output_tokens": 448,
     "cost_usd_list": 0.0226139,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q10",
   "type": "cross",
   "question": "What end dates do the papers give for the LBK in the Southern Buh/Dniester region, at Vráble, in the Rhineland-Limburg-Hesbaye corridor (Bayesian Model 6) and in the Alsace Plain?",
   "ground_truth": "Southern Buh/Dniester (Kiosak & Kotova): end of LBK 5207-5048 cal BC (68.3%) or 5211-5010 cal BC (95.4%) in the overlapping-phases model. Vráble: occupation ends around 5000-4950 cal BCE. Rhineland-Limburg-Hesbaye (Kroon et al., Model 6): latest possible end 5042-4986 BC (68.3%) or 5069-4939 BC (95.4%). Alsace Plain (Waldvogel): LBK occupation between 5300 and 4950 BC.",
   "key_facts": [
    "Buh/Dniester 5207-5048 (68.3%) / 5211-5010 (95.4%)",
    "Vráble end around 5000-4950 cal BCE",
    "Kroon Model 6 5042-4986 (68.3%) / 5069-4939 (95.4%)",
    "Alsace until 4950 BC"
   ],
   "approaches": {
    "none": {
     "answer": "I can't reliably give the end dates those papers report, and I don't want to guess at figures I can't tie to the source.\n\n- **Vráble (Slovakia):** My recollection is that the settlement ran from roughly 5250 to 4950 cal BC (Hofmann et al. and colleagues). I'm not certain of the exact modelled end boundary.\n- **Alsace:** I vaguely recall the LBK ending around 4900 cal BC, followed by the Bischheim/Hinkelstein-type transition (Denaire et al.). I can't confirm the paper's exact range.\n- **Southern Buh/Dniester and Rhineland-Limburg-Hesbaye (Bayesian Model 6):** I don't know the end dates the papers give.\n\nIf you can share the papers or the relevant passages, I can help extract and compare the dates.",
     "verdict": "partial",
     "facts_found": 1,
     "facts_total": 4,
     "judge_note": "The Vráble end of about 4950 cal BC roughly matches, but the Alsace end (~4900 instead of 4950 BC) is off, and the Buh/Dniester and Kroon Model 6 dates are not given.",
     "wall_seconds": 7.04,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 716,
     "cache_read_tokens": 0,
     "output_tokens": 572,
     "cost_usd_list": 0.008588,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The excerpts don't give end dates for the Southern Buh/Dniester region, Vráble, the Alsace Plain, or for Model 6. They never mention the first two. Alsace appears only as \"Lower Alsace\", and only for the start of the LBK, which is roughly contemporaneous with the Rhineland start in the last decade of the 54th century BC. Model 6 is described only as shortening durations, especially for early chronozones, with no end dates.\n\nThe excerpts do give end dates from Models 8-10 for part of the corridor:\n\n- **Limburg (Model 9):** latest LBK occupation about 5107-5013 BC (68.3%), possibly 5125-4974 BC (95.4%). This is uncertain because the youngest dated site, Maastricht-Cannerberg, is an outlier in the model.\n- **Rhineland (Model 8):** latest LBK 5186-4871 BC (68.3%), possibly 5199-4782 BC (95.4%).\n- **Hesbaye (Model 10):** the end date is not given in these excerpts.",
     "verdict": "abstained",
     "facts_found": 0,
     "facts_total": 4,
     "judge_note": "The answer says none of the four requested end dates are reported and gives figures only from Models 8-10, which match no key fact.",
     "wall_seconds": 6.3,
     "retrieval_seconds": 1.91,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6836,
     "cache_read_tokens": 0,
     "output_tokens": 626,
     "cost_usd_list": 0.033608,
     "retrieved_chunks": [
      "W7225183261#30",
      "W7225183261#25",
      "W7225183261#26",
      "W7225183261#33",
      "W7225183261#35",
      "W7225183261#28",
      "W7225183261#0",
      "W7225183261#31"
     ]
    },
    "okf": {
     "answer": "These are the end dates the bundle gives for each region.\n\n- **Southern Buh/Dniester:** The modelled end of the LBK is 5207–5048 cal BC at 68.3% and 5211–5010 cal BC at 95.4%. This range covers the late-6th-millennium radiocarbon plateau.\n- **Vráble:** The settlement ends c. 5000–4950 cal BCE. The site is dated roughly 5250–4950 BCE (Meadows et al. 2019). The mass deposition of headless skeletons dates to 5300/5210–5050/4950 cal BCE.\n- **Rhineland–Limburg–Hesbaye, Bayesian Model 6:** The latest LBK end is 5042–4986 BC at 68.3% and 5069–4939 BC at 95.4%.\n- **Alsace Plain:** The LBK ends c. 4950 BC (the LBK spans c. 5300–4950 BC, phases II–V). The bundle notes a hiatus after the LBK before Hinkelstein. Hinkelstein reached Lower Alsace in 4910–4725 BC (68% probability, Denaire et al. 2017).",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "All four end dates match the ground truth; the extra details do not contradict it.",
     "wall_seconds": 17.05,
     "retrieval_seconds": null,
     "turns": 10,
     "input_tokens": 10,
     "cache_creation_tokens": 18237,
     "cache_read_tokens": 41722,
     "output_tokens": 1898,
     "cost_usd_list": 0.09612019999999999,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "**Southern Buh/Dniester (Kiosak & Kotova):** Bayesian modelling puts the end of the LBK at 5207–5048 cal BC (68.3%) or 5211–5010 cal BC (95.4%) in the overlapping-phases model. The sequential model gives an LBK span of 5213–5050 cal BC (95.4%), ending 5125–5046 (68.3%). A gap of about 300–400 years follows before Early Trypillia.\n\n**Vráble (Furholt et al.):** Occupation of the site ends around 5000–4950 cal BCE. The mass deposition of headless bodies dates anywhere between 5210 cal BCE and the end of occupation. The late LBK crisis horizon at enclosures is dated 5100–5000 BCE.\n\n**Rhineland–Limburg–Hesbaye, Model 6 (Kroon et al.):** The latest possible end of LBK inhabitation is 5042–4986 BC (68.3%) or 5069–4939 BC (95.4%). The earliest start is 5176–5082 BC (68.3%).\n\n**Alsace Plain (Waldvogel):** LBK occupation runs c. 5300–4950 BC, ending around 4950 BC. Denaire et al. (2017) place the Hinkelstein arrival at 4910–4725 BC, 40–175 years after the LBK ended.",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "All four end dates match the ground truth; the extra details do not contradict it.",
     "wall_seconds": 5.25,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 616,
     "cache_read_tokens": 158219,
     "output_tokens": 462,
     "cost_usd_list": 0.0229099,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q11",
   "type": "cross",
   "question": "Which papers discuss the radiocarbon calibration curve plateau around the end of the LBK, and what do they conclude about it?",
   "ground_truth": "Furholt et al. (Vráble): a flat section of the calibration curve between 5210 and 5000 cal BCE prevents a more precise chronology of the mass deposition and also hinders testing LBK chronologies with absolute dates. Kiosak & Kotova: the end of the LBK falls on the 'infamous' late 6th-millennium BC plateau, which stretches the calibration of some dates (such as TKA-20833). Kroon et al.: considered a calibration-curve plateau as one possible cause of the poor model fit, but a simulation with simulated dates (Model 5) showed that neither the model structure nor the shape of IntCal20 explains it; the problem lies in a mismatch between dates and the overarching typochronology.",
   "key_facts": [
    "Vráble: plateau 5210-5000 cal BCE hinders precision",
    "Buh/Dniester: late 6th millennium plateau stretches calibration of dates",
    "Kroon: plateau tested via simulation and ruled out as the explanation"
   ],
   "approaches": {
    "none": {
     "answer": "I can't reliably name specific papers that address a calibration plateau at the end of the LBK, and I'd rather not invent citations or conclusions.\n\nWhat I'm fairly confident of:\n\n- The LBK ends around 5000 BC (roughly 5100-4900 cal BC, varying by region). Flat or wiggly stretches of the curve in this period widen calibrated date ranges and make it hard to resolve the order and duration of the final LBK phases and the transition to successors like Hinkelstein, Großgartach and Bischheim.\n- The usual response in the literature is Bayesian chronological modelling with stratigraphic and typological priors. Denaire et al. 2017 (*Proceedings of the Prehistoric Society*), on the Neolithic of Alsace and the Rhine region, is one example I recall, though I can't vouch for what it says specifically about a plateau.\n- Work on the \"end of the LBK\" crisis, such as the Talheim, Schletz and Asparn massacres, and Gronenborn's climate-related arguments, depends on such tight dating.\n\nFor a reliable answer, I'd search for \"LBK end chronology Bayesian IntCal20 plateau\", and check Denaire, Jeunesse, Lefranc and Gronenborn, plus IntCal20 papers.",
     "verdict": "abstained",
     "facts_found": 0,
     "facts_total": 3,
     "judge_note": "The answer declines to name the papers and offers only general background, with none of the specific findings from Vráble, Buh/Dniester or Kroon.",
     "wall_seconds": 12.68,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 678,
     "cache_read_tokens": 0,
     "output_tokens": 1111,
     "cost_usd_list": 0.013826000000000001,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "Two of the excerpted papers discuss the plateau.\n\n**W7204009876 (Kiosak – Kotova, post-LBK hunter-gatherers between the Southern Buh and Dniester):** The Bayesian model puts the end of LBK at 5207–5048 cal BC (68.3%) or 5211–5010 cal BC (95.4%). It notes this \"basically covers all of the infamous late 6th-millennium BC radiocarbon plateau\" (Stäuble 2013). The start of the para-Neolithic phase gets very similar estimates (5202–5066 and 5317–5048 cal BC). The authors attribute the apparent overlap between LBK and the hunter-gatherer dates partly to the \"plateau effect\": some dates, such as TKA-20833, fell partly into the plateau, so their calibration was stretched. They conclude that direct evidence for contemporaneity is limited. For example, at Bazkiv Ostriv, a food-crust date from the same vessel falls earlier, within the LBK period, so contemporaneity is statistically unlikely.\n\n**W7225183261 (Kroon et al., LBK/post-LBK in Hesbaye, Limburg, Rhineland):** This paper discusses the calibration curve's shape only as a possible cause of a poor model fit (Amodel = 7.4%). Simulated dates (Model 5) fit well (Amodel = 67.5%). It concludes that neither the model structure nor the IntCal20 curve shape explains the poor fit. The cause is more likely a mismatch between the radiocarbon dates and the typochronology. It does not mention the plateau by name in these excerpts.",
     "verdict": "partial",
     "facts_found": 2,
     "facts_total": 3,
     "judge_note": "Covers the Kiosak-Kotova plateau and the Kroon simulation result correctly, but leaves out the Vráble 5210-5000 cal BCE plateau entirely.",
     "wall_seconds": 5.33,
     "retrieval_seconds": 1.87,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6800,
     "cache_read_tokens": 0,
     "output_tokens": 548,
     "cost_usd_list": 0.032684,
     "retrieved_chunks": [
      "W7204009876#15",
      "W7225183261#25",
      "W7225183261#30",
      "W7225183261#18",
      "W7204009876#3",
      "W7225183261#17",
      "W7204009876#13",
      "W7225183261#33"
     ]
    },
    "okf": {
     "answer": "Two papers in the bundle discuss the plateau. Both treat it as a flat stretch of the calibration curve at c. 5210–5000 cal BC. Both say it limits how finely the end of the LBK can be dated.\n\n1. **\"Neolithic Bodies in Vráble – 7000 year-old Headless Human Skeletons in an Enclosed LBK Settlement in South-West Slovakia\" (W7163143446)**\n   - The skeletons date to 5300/5210–5050/4950 cal BCE.\n   - The plateau blocks a finer chronology. The deposition was probably a few years or less, but it can't be settled whether it was a one-off event or a recurring long-term practice.\n   - The plateau also hinders absolute control of pottery-based LBK chronologies.\n\n2. **\"The issue of the 'resurgence' of hunter-gatherer lifestyles in the region between the middle courses of the Southern Buh and Dniester rivers during the post-LBK period\" (W7204009876)**\n   - The Bayesian model puts the end of the LBK at 5211–5010 cal BC (95.4%), which basically covers the \"infamous\" plateau.\n   - Some dates, such as TKA-20833, partly fell into the plateau, so their calibration was stretched to its limits.\n   - The paper lists the plateau as one of the factors behind a blurry radiocarbon chronology.\n\nThe Vráble finding also cites Masclans et al. 2025 and Morell-Rovira et al. 2025 on the plateau. The bundle has no separate entries for them.",
     "verdict": "partial",
     "facts_found": 2,
     "facts_total": 3,
     "judge_note": "Vráble and Buh/Dniester are covered correctly, but Kroon et al.'s simulation that ruled out the plateau as the explanation is missing.",
     "wall_seconds": 10.36,
     "retrieval_seconds": null,
     "turns": 5,
     "input_tokens": 6,
     "cache_creation_tokens": 3340,
     "cache_read_tokens": 11636,
     "output_tokens": 1181,
     "cost_usd_list": 0.026345600000000004,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "Three of the four papers touch on the calibration plateau; Waldvogel (Alsace funerary variability) and Juźwińska et al. (Polish LBK plant remains) do not discuss it.\n\n**Furholt et al. (Vráble, Slovakia)**\n- They say a flat section of the calibration curve between 5210 and 5000 cal BCE has prevented a more precise radiocarbon chronology for the headless-skeleton mass deposition.\n- The deposition could fall anywhere between 5210 cal BCE and the end of occupation around 5000–4950 cal BCE.\n- The same plateau stops them deciding whether head removal was recurring or a one-off event. It also prevents absolute-date checks on pottery-based LBK chronologies, so the idea that these events mark the final LBK phase is \"not yet well established\".\n\n**Kiosak & Kotova (Southern Buh/Dniester)**\n- Their model puts the end of the LBK at 5207–5048 cal BC (68.3%), which covers essentially the whole \"infamous\" late 6th-millennium plateau (citing Stäuble 2013).\n- The plateau stretches the calibration of some dates, such as TKA-20833, and makes the hunter-gatherer dates appear to overlap with the LBK.\n- They conclude that direct evidence for contemporaneity is limited.\n\n**Kroon et al. (Rhineland–Limburg–Hesbaye)**\n- They list plateaus in the calibration curve as one possible cause of poor model agreement.\n- A simulated-dates test (Model 5) shows that neither the model structure nor the IntCal20 curve explains the problem. They attribute it to a mismatch between the radiocarbon dates and the chronozones.\n- They add that a small set of precise AMS dates could substantiate the typochronology.\n\nAll three conclude that the plateau limits resolution and that more precise dating is needed.",
     "verdict": "partial",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "All three key facts are present, but the closing claim that all three papers see the plateau as limiting resolution contradicts Kroon et al., who ruled it out (and the intro miscounts the papers).",
     "wall_seconds": 8.16,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 578,
     "cache_read_tokens": 158219,
     "output_tokens": 854,
     "cost_usd_list": 0.026677900000000004,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q12",
   "type": "cross",
   "question": "When does the LBK first appear in southeastern Poland, in the Alsace Plain, in Dutch Limburg and in the Rhineland according to the papers?",
   "ground_truth": "Southeastern Poland: around 5300 bc, during the Pre-Music Note phase. Alsace Plain: 5300 BC (LBK phases II to V, until 4950 BC). Dutch Limburg: around 5250 BC (Modderman Ib). Rhineland: late 54th century BC (house generation I).",
   "key_facts": [
    "SE Poland ~5300 (Pre-Music Note)",
    "Alsace 5300 BC",
    "Limburg ~5250 BC (Modderman Ib)",
    "Rhineland late 54th century BC"
   ],
   "approaches": {
    "none": {
     "answer": "I don't have the papers you mean, so I can't say what they report. From general knowledge, the LBK spread from Transdanubia (Hungary), starting around 5500 BC, and reached these regions at roughly the same time:\n\n- **Southeastern Poland (Lesser Poland):** about 5300 BC, possibly slightly earlier (~5400-5300 BC), according to some estimates.\n- **Alsace Plain:** about 5300 BC. Some Bayesian models put the earliest sites slightly later, around 5200 BC.\n- **Dutch Limburg:** about 5300 BC.\n- **Rhineland:** about 5300 BC.\n\nThese dates are approximate, and I'm not confident of the exact figures or which models give which. If you share the papers or their dates, I can compare them.",
     "verdict": "partial",
     "facts_found": 3,
     "facts_total": 4,
     "judge_note": "The SE Poland, Alsace and Rhineland dates roughly match, but Dutch Limburg is given as ~5300 BC instead of ~5250 BC (Modderman Ib), and the answer is general knowledge rather than drawn from the papers.",
     "wall_seconds": 8.58,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 688,
     "cache_read_tokens": 0,
     "output_tokens": 767,
     "cost_usd_list": 0.010426,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "**Southeastern Poland:** The excerpt only says that farming in southern Poland began in the latter half of the 6th millennium BC, with the arrival of the first LBK groups from the Danube region. It gives no more precise date for the southeast.\n\n**Alsace Plain:** The excerpts give no explicit date. They only say that the Rhineland's earliest LBK appearance is roughly contemporaneous with Denaire et al. (2017)'s estimates for Lower Alsace. That puts Lower Alsace at about the end of the 54th century BC. Alsace is not covered in more detail here.\n\n**Dutch Limburg:** The models place the earliest LBK at 5171–5091 BC (68.3% probability), possibly 5218–5066 BC (95.4%). This falls in the 52nd century BC and may coincide with Hesbaye.\n\n**Rhineland:** The earliest LBK most likely appeared in the last decade of the 54th century BC. The excerpts give no exact range.\n\nThis implies a gap of at least one century between the Rhineland and Limburg.",
     "verdict": "partial",
     "facts_found": 2,
     "facts_total": 4,
     "judge_note": "Rhineland (late 54th century BC) and Alsace (about the end of the 54th century, roughly 5300 BC) match. SE Poland lacks the ~5300 BC / Pre-Music Note date, and Limburg is given as 5171–5091 BC instead of ~5250 BC (Modderman Ib).",
     "wall_seconds": 6.54,
     "retrieval_seconds": 2.07,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6818,
     "cache_read_tokens": 0,
     "output_tokens": 640,
     "cost_usd_list": 0.033676,
     "retrieved_chunks": [
      "W7225183261#28",
      "W7225183261#38",
      "W7225183261#35",
      "W7225183261#3",
      "W7225183261#30",
      "W7225183261#8",
      "W7225183261#11",
      "W7166088190#0"
     ]
    },
    "okf": {
     "answer": "According to the papers in the bundle:\n\n- **Southeastern Poland:** the LBK appears around 5300 BC, in the Pre-Music Note phase (Juźwińska et al. 2026).\n- **Alsace Plain:** the LBK dates to c. 5300–4950 BC (phases II–V), so it starts around 5300 BC (Waldvogel 2026). The paper also notes that recent finds at Ensisheim, from phase II, suggest Facies I and II did not emerge before phase III.\n- **Rhineland:** the modelled start is 5460–5310 BC at 68.3% probability, or 5550–5224 BC at 95.4%. The authors think arrival was most likely in the last decade of the 54th century BC (Kroon et al. 2026).\n- **Dutch Limburg:** the modelled start is 5171–5091 BC at 68.3%, or 5218–5066 BC at 95.4%. The authors place arrival in the 52nd century BC (Kroon et al. 2026).\n\nThe Rhineland therefore has the earliest start, and Limburg is roughly two centuries later.",
     "verdict": "partial",
     "facts_found": 3,
     "facts_total": 4,
     "judge_note": "SE Poland, Alsace and Rhineland match the ground truth, but Limburg is given as the 52nd century (5171–5091 BC) rather than c. 5250 BC (Modderman Ib).",
     "wall_seconds": 10.7,
     "retrieval_seconds": null,
     "turns": 7,
     "input_tokens": 8,
     "cache_creation_tokens": 8284,
     "cache_read_tokens": 23737,
     "output_tokens": 1143,
     "cost_usd_list": 0.046955699999999996,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "**Southeastern Poland** (Juźwińska et al.): LBK appears around 5300 BC, in the Pre-Music Note phase, spreading along the Wisłok and then San valleys. (More generally, LBK reached southern Poland via Moravia and Slovakia around 5350 BC.)\n\n**Alsace Plain** (Waldvogel): LBK communities occupied it from 5300 BC (phases II–V) until about 4950 BC.\n\n**Dutch Limburg** (Kroon et al.): Typochronology/older literature dates initial occupation to c. 5250 BC (Modderman Ib). The Bayesian models (Models 9–10) place the earliest arrival later, at 5171–5091 BC (68.3%), or 5218–5066 BC (95.4%), i.e. the 52nd century BC. This coincides with Hesbaye (5193–5087 BC).\n\n**Rhineland** (Kroon et al.): Conventional view is the late 54th century BC (HG I). The models give the earliest arrival at 5460–5310 BC (68.3%), possibly 5550–5224 BC (95.4%), most likely in the last decade of the 54th century BC. The Rhineland is thus the earliest, with a gap of at least a century before Limburg and Hesbaye.",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "All four key dates are present; the extra Bayesian model ranges are reported as additional findings and do not override the conventional dates in the ground truth.",
     "wall_seconds": 5.03,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 588,
     "cache_read_tokens": 158219,
     "output_tokens": 429,
     "cost_usd_list": 0.0224679,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q13",
   "type": "cross",
   "question": "Which papers discuss hunter-gatherer or Mesolithic-descended groups living alongside or after the LBK farmers, and what is the evidence?",
   "ground_truth": "Kiosak & Kotova (Southern Buh/Dniester): ceramic hunter-gatherers ('para-Neolithic'); 69 radiocarbon dates give a bimodal distribution with a pre-LBK peak, a decline during LBK expansion and a renewed peak after the LBK collapse, which the authors read as a hunter-gatherer revival before the Early Trypillian expansion. Kroon et al. (Rhineland-Limburg-Hesbaye): Swifterbant communities with strong ties to Late Mesolithic groups lived in the wetlands of the Netherlands and adjacent Belgium; they interacted with early farmers and integrated domestic animals (as of 4450 BC) and crops (as of 4200 BC); Mesolithic populations are continuous in Limburg.",
   "key_facts": [
    "Buh/Dniester: para-Neolithic ceramic hunter-gatherers, second radiocarbon peak after LBK",
    "Kroon: Swifterbant with Late Mesolithic ties",
    "domestic animals from 4450 BC, crops from 4200 BC"
   ],
   "approaches": {
    "none": {
     "answer": "I'm working from memory, with no database access, so verify details before citing.\n\n**Key papers**\n- **Bollongino et al. 2013, *Science*** (\"2000 years of parallel societies\"): At Blätterhöhle (Germany), mtDNA shows hunter-gatherer lineages (haplogroup U) persisting for about 2,000 years alongside farmer groups, with different diets (including more fish) and little mixing at first.\n- **Haak et al. 2015, *Nature*, and Lipson et al. 2017, *Nature***: Early LBK farmers carry almost no hunter-gatherer ancestry. It rises in later Neolithic farmers, which implies later admixture with local foragers.\n- **Rivollat et al. 2020, *Science Advances***: In France, Mesolithic-related ancestry in farmers shows interaction and admixture with local foragers.\n- **Mathieson et al. 2018, *Nature***: In the Iron Gates region, foragers and farmers mixed.\n- **Archaeological work (Gronenborn, Lüning, Zvelebil)**: Hunter-gatherer contact is inferred from La Hoguette and Limburg pottery, and from Mesolithic sites that persisted in marginal areas.",
     "verdict": "wrong",
     "facts_found": 0,
     "facts_total": 3,
     "judge_note": "The answer cites other real studies from memory but none of the key facts: no Buh/Dniester para-Neolithic radiocarbon peaks, no Kroon/Swifterbant, and no 4450/4200 BC dates.",
     "wall_seconds": 14.72,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 686,
     "cache_read_tokens": 0,
     "output_tokens": 1498,
     "cost_usd_list": 0.017728,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "Two papers in the excerpts address this.\n\n**Kiosak & Kotova (W7204009876), Southern Buh-Dniester region:**\n- Bayesian analysis shows a gap of about 300-400 years between the LBK decline and the arrival of Early Trypillia. Several early 5th-millennium BC radiocarbon dates within it come from riverside camps of ceramic hunter-gatherers (para-Neolithic) with Savran-style pottery.\n- LBK-era dates drop markedly, and hundreds of LBK settlements contrast with only a few dozen forager camps.\n- Early Trypillian sites show foraging traces: syncretic vessels, high proportions of wild fauna, fishing, bone and antler industries, and microburin rhomboid points.\n- Coexistence with the LBK itself is not directly proven, and the authors say no single model can yet be accepted or rejected.\n\n**Settlement and temporal dynamics (W7225183261), Rhineland, Hesbaye, Limburg:**\n- It cites other work (Amkreutz 2013; Olalde et al. 2026) saying that \"SWB\" communities with strong ties to Mesolithic populations co-existed with and persisted after LBK communities in the Netherlands.\n- The excerpt gives no direct evidence for this beyond the citations.",
     "verdict": "partial",
     "facts_found": 2,
     "facts_total": 3,
     "judge_note": "The answer covers the para-Neolithic hunter-gatherers with post-LBK dates and the Swifterbant groups' Mesolithic ties, but omits the 4450 BC domestic-animal and 4200 BC crop integration dates.",
     "wall_seconds": 7.5,
     "retrieval_seconds": 2.03,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6987,
     "cache_read_tokens": 0,
     "output_tokens": 848,
     "cost_usd_list": 0.036432,
     "retrieved_chunks": [
      "W7204009876#0",
      "W7204009876#20",
      "W7204009876#21",
      "W7204009876#19",
      "W7204009876#22",
      "W7225183261#3",
      "W7204009876#26",
      "W7204009876#3"
     ]
    },
    "okf": {
     "answer": "Two papers in the bundle cover this.\n\n**1. Kiosak & Kotova 2026, Archeologické rozhledy 78 (Southern Buh–Dniester region)**\nThis paper argues for a \"resurgence\" of ceramic (para-Neolithic) hunter-gatherers after the LBK. It does not find clear evidence that they lived alongside LBK farmers.\n- **Dates:** 69 radiocarbon dates give a bimodal kernel density estimate. There is a pre-LBK peak, a decline during the LBK, and a second peak after the LBK collapse. The authors read this as evidence against unaffected coexistence.\n- **Contact:** LBK fineware occurs in layers with forager pottery at Bazkiv Ostriv, Gyrzheve, Soroca 5, Tătărăuca Nouă 15 and others. These are not closed contexts and may be mechanical admixtures. No direct, indisputable evidence of coexistence was found. R_Combine makes coexistence at Bazkiv Ostriv statistically unlikely.\n- **After the LBK:** Savran-style pottery has northern and eastern elements, which suggests non-local groups. Early Trypillian sites also show forager traits: syncretic vessels, wild fauna, fishing, and microburin points.\n- **Alternatives left open:** prolonged coexistence, former LBK farmers turning to foraging, and continuous local development into Early Trypillia.\n\n**2. Kroon et al. 2026 (Rhineland–Hesbaye loess corridor)**\nThis paper uses Bayesian modelling to date the Swifterbant culture, a Late Mesolithic-descended group in the Dutch wetlands. Swifterbant first appears at 5542–5504 BC (68.3%) and overlaps the LBK for about 426–515 years. It persisted through the LBK and post-LBK periods.",
     "verdict": "partial",
     "facts_found": 2,
     "facts_total": 3,
     "judge_note": "Covers the para-Neolithic hunter-gatherers with a second post-LBK radiocarbon peak and Swifterbant's Late Mesolithic ties, but omits domestic animals from 4450 BC and crops from 4200 BC.",
     "wall_seconds": 13.61,
     "retrieval_seconds": null,
     "turns": 9,
     "input_tokens": 6,
     "cache_creation_tokens": 7472,
     "cache_read_tokens": 10799,
     "output_tokens": 1630,
     "cost_usd_list": 0.0472799,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "Two papers address this directly, and a third only touches on it.\n\n**Kiosak & Kotova (Southern Buh/Dniester, Ukraine):** They argue for a hunter-gatherer \"resurgence\" after the LBK.\n- Bayesian modelling shows a gap of about 371-602 years (95.4%) between the end of the LBK and the arrival of Early Trypillia.\n- Of 69 hunter-gatherer radiocarbon dates, kernel density estimation gives a bimodal pattern: a pre-LBK peak, a dip during the LBK, and a second peak after it.\n- The second-peak dates come from riverside camps with Savran-style pottery: Mykolyna Broiaka, Puhach 2, Shumyliv-Cherniatka, Bazkiv Ostriv, Zakharivka I, Melnychna Krucha R4 and Tătărăuca Nouă 15.\n- Some vessel forms and decoration parallel Azov-Dnieper and Kyiv-Cherkassy pottery to the east and north.\n- Coexistence with the LBK is only possible, not shown, because the evidence comes from mixed layers or unreliable dates.\n\n**Kroon et al. (Rhineland-Limburg-Hesbaye):** Swifterbant communities with strong Late Mesolithic ties persisted alongside and after the LBK in the Netherlands and Belgium.\n- Model 9 puts their start at 5542-5504 BC (68.3%), with 426-515 years of overlap with the LBK.\n- They took up domestic animals around 4450 BC and crops around 4200 BC.\n- Limburg was apparently abandoned by farmers, while Mesolithic populations continued.\n\n**Waldvogel (Alsace):** This paper only touches on the topic. It notes that some Upper Alsace LBK graves have ornaments (marine-shell beads) suggesting close links with Mesolithic populations.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer covers the pottery-using hunter-gatherers with a second radiocarbon peak after the LBK, Swifterbant's Late Mesolithic ties, and the 4450/4200 BC dates; extra details do not contradict the ground truth.",
     "wall_seconds": 10.99,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 586,
     "cache_read_tokens": 158219,
     "output_tokens": 1268,
     "cost_usd_list": 0.0308499,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q14",
   "type": "cross",
   "question": "Which post-LBK cultures are discussed for the Alsace Plain and for the Rhineland-Limburg-Hesbaye corridor, and which appear in both?",
   "ground_truth": "Alsace Plain (Waldvogel): the Hinkelstein, Grossgartach and Rössen sequence. Rhineland-Limburg-Hesbaye (Kroon et al.): Grossgartach (GG), Rössen (RÖ), Blicquy/Villeneuve-Saint-Germain (BVSG) and Swifterbant (SWB). Grossgartach and Rössen appear in both.",
   "key_facts": [
    "Alsace: Hinkelstein, Grossgartach, Rössen",
    "Kroon: GG, RÖ, BVSG, SWB",
    "shared: Grossgartach and Rössen"
   ],
   "approaches": {
    "none": {
     "answer": "I can't tell which source you mean, so this is a general answer from my own knowledge and may not match its exact wording.\n\n**Alsace Plain (Upper Rhine):** The usual sequence is Hinkelstein, then Grossgartach, then Rössen, with Bischheim and Planig-Friedberg in the regional literature. Michelsberg and Munzingen follow later.\n\n**Rhineland-Limburg-Hesbaye corridor:** Here the discussion usually centres on Rössen, along with Blicquy/Villeneuve-Saint-Germain on the western edge, and later Michelsberg. Hinkelstein-type influences are sometimes mentioned in the Rhineland.\n\n**Appearing in both:** Rössen is the clearest overlap, and Michelsberg is the other likely one, as a later Neolithic culture present in both regions.\n\nIf you can name the paper or share its text, I can check this against what it actually says.",
     "verdict": "partial",
     "facts_found": 1,
     "facts_total": 3,
     "judge_note": "Alsace sequence is correct, but for the corridor it omits Grossgartach and Swifterbant, and it lists Michelsberg as shared instead of Grossgartach.",
     "wall_seconds": 6.83,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 691,
     "cache_read_tokens": 0,
     "output_tokens": 638,
     "cost_usd_list": 0.009148,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The excerpts don't mention the Alsace Plain, so I can't say which post-LBK cultures apply there or which appear in both areas.\n\nFor the Rhineland-Limburg-Hesbaye corridor, the excerpts name these post-LBK groups:\n- **Rhineland:** Grossgartach (GG). An occupation gap separates it from the LBK, though the models can't confirm this because the post-LBK dates are sparse.\n- **Limburg:** RÖ communities, after a settlement hiatus. The authors found no suitable dates for RÖ in the Netherlands. SWB societies with ties to Mesolithic populations persisted through the LBK and post-LBK period.\n- **Hesbaye:** Blicquy/Villeneuve-Saint-Germain (BVSG), which was continuous with the LBK, or possibly partly contemporaneous with it. Its initial phase may have emerged in Hainaut.",
     "verdict": "partial",
     "facts_found": 1,
     "facts_total": 3,
     "judge_note": "Lists all four Kroon cultures (GG, RÖ, BVSG, SWB) but says Alsace is not covered, so it misses the Hinkelstein-Grossgartach-Rössen sequence and the shared cultures.",
     "wall_seconds": 6.03,
     "retrieval_seconds": 1.93,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6967,
     "cache_read_tokens": 0,
     "output_tokens": 527,
     "cost_usd_list": 0.033142,
     "retrieved_chunks": [
      "W7225183261#38",
      "W7225183261#0",
      "W7225183261#11",
      "W7225183261#12",
      "W7225183261#29",
      "W7225183261#3",
      "W7225183261#28",
      "W7225183261#30"
     ]
    },
    "okf": {
     "answer": "**Alsace Plain** (Waldvogel 2026): the post-LBK cultures are Hinkelstein (c. 4800–4730 BC, only 2 graves), Grossgartach (c. 4730–4600 BC, 228 inhumations) and Rössen (c. 4600–4400 BC, 22 inhumations). The bundle also describes Villeneuve-Saint-Germain (VSG) in the Paris Basin as part of the Alsace funerary comparison, but not as an Alsace Plain culture.\n\n**Rhineland–Limburg–Hesbaye corridor** (Kroon et al. 2026): the post-LBK groups are Großgartach (GG), Rössen (RÖ), Blicquy/Villeneuve-Saint-Germain (BVSG), and Swifterbant (SWB) as a wetland, Mesolithic-linked group.\n- Rhineland: GG from Middle GG (c. 4900–4850 BC) and RÖ (4750–4600 BC).\n- Limburg: RÖ near Maastricht-Randwijck.\n- Hesbaye: BVSG.\n\n**In both:** Grossgartach and Rössen. Hinkelstein is only in the Alsace paper. BVSG and Swifterbant are only in the corridor paper. The VSG/BVSG name is shared across the two papers, but they apply it to different regions.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer lists all Alsace and Kroon cultures and correctly names Grossgartach and Rössen as shared; the extra VSG detail does not contradict the ground truth.",
     "wall_seconds": 12.45,
     "retrieval_seconds": null,
     "turns": 7,
     "input_tokens": 8,
     "cache_creation_tokens": 11166,
     "cache_read_tokens": 25678,
     "output_tokens": 1346,
     "cost_usd_list": 0.0607078,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "**Alsace Plain (Waldvogel):** Hinkelstein (HST, c. 4800–4730 BC; only two burials), Grossgartach (GG, c. 4730–4600 BC; 228 inhumations) and Rössen (RO, c. 4600–4400 BC; 22 inhumations).\n\n**Rhineland–Limburg–Hesbaye corridor (Kroon et al.):** Großgartach (GG), Rössen (RÖ), Blicquy/Villeneuve-Saint-Germain (BVSG), and Swifterbant (SWB), the last tied to Late Mesolithic groups. In the Rhineland, GG arrives in the Middle GG phase (4900–4850 BC) and RÖ follows (4750–4600 BC). In Limburg, only one Rössen settlement is known (Maastricht-Randwijck). Hesbaye has two BVSG settlements.\n\n**Appearing in both:** Grossgartach and Rössen.\n\nHinkelstein is discussed only for Alsace, whereas BVSG and Swifterbant are discussed only for the corridor. Waldvogel does mention Villeneuve-Saint-Germain in passing, as a continuation of Paris Basin funerary practice (sub-facies Ib), but it is not a post-LBK culture of the Alsace Plain itself.",
     "verdict": "correct",
     "facts_found": 3,
     "facts_total": 3,
     "judge_note": "The answer gives the full Alsace sequence (Hinkelstein, Grossgartach, Rössen), all four Kroon cultures (GG, RÖ, BVSG, SWB), and correctly names Grossgartach and Rössen as the shared cultures.",
     "wall_seconds": 4.77,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 591,
     "cache_read_tokens": 158219,
     "output_tokens": 450,
     "cost_usd_list": 0.022689900000000002,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q15",
   "type": "synthesis",
   "question": "What is the story with the headless skeletons found at the end of the LBK?",
   "ground_truth": "At Vráble (Slovakia), a mass deposition of at least 77 headless skeletons was found in the outer ditch of an enclosed LBK settlement, dated to about 5300/5210-5050/4950 cal BCE; the heads are missing (cutmarks suggest cutting with sharp tools) and little else is known, so the motive is open. Interpretations discussed: headhunting, massacre (made less likely by patterned placements elsewhere), human sacrifice or a special burial rite; the authors put it in a context of similar late-LBK ditch depositions (Herxheim, Asparn-Schletz, Talheim and others) and argue against explaining it as modern-style war over land, preferring an immanentist cosmology. Aspects such as aDNA, isotopes and kinship are still under analysis.",
   "key_facts": [
    "Vráble, at least 77 headless skeletons in the ditch",
    "heads missing",
    "interpretations open: headhunting, massacre, sacrifice, burial rite",
    "compared with Herxheim, Asparn-Schletz, Talheim",
    "analyses still under way"
   ],
   "approaches": {
    "none": {
     "answer": "I can't pin down one specific \"headless skeleton\" find, so I'll say what I'm confident about. The late LBK (c. 5000 BCE) has several sites with violence or unusual treatment of the dead:\n\n- **Herxheim (Germany, c. 5300-4950 BCE):** An enclosure ditch held the remains of at least 500 individuals, probably many more. Bodies were dismembered and defleshed, and bones were broken. Skulls were often selected and modified into skullcaps, and many were deposited without bodies. Zeeb-Lanz and colleagues see a large-scale ritual; Boulestin and others argue for cannibalism. Both agree many victims were non-local.\n- **Talheim (Germany):** Around 34 men, women and children were killed with adze and club blows and thrown into a pit. This was a massacre, not a skull ritual.\n- **Schletz-Asparn (Austria):** Many skeletons were found in the enclosure ditch, with signs of violence and some incomplete bodies.\n\nIf you mean a particular site, tell me and I'll narrow it down.",
     "verdict": "partial",
     "facts_found": 1,
     "facts_total": 5,
     "judge_note": "The answer does not mention Vráble, the 77 headless skeletons, the range of interpretations or the ongoing analyses; it only covers the comparison sites Herxheim, Talheim and Asparn-Schletz.",
     "wall_seconds": 11.67,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 659,
     "cache_read_tokens": 0,
     "output_tokens": 1041,
     "cost_usd_list": 0.01305,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The excerpts describe a mass deposition of headless skeletons at the enclosed LBK settlement of Vráble-Veľké Lehemby, south-west Slovakia. They don't say anything specific about it falling at the end of the LBK.\n\n- **Scale:** The deposit lies in the outer ditch terminal east of entrance 2. It was first reported in 2022 as 37 skeletons, 36 of them headless. By 2024 it stood at at least 77 headless skeletons and one complete skeleton within 25 m, and its eastern end hasn't been reached. Forty-four lay in the first 9.5 m.\n- **Arrangement:** The bodies were prone, supine or twisted, with no consistent orientation, and sometimes overlay each other.\n- **Timing:** Well-aligned neck vertebrae and well-preserved hands and feet suggest the bodies were not far into decomposition when deposited. Some were placed after the heads were removed.\n- **Head removal:** Preliminary cutmarks on cervical vertebrae suggest cutting with sharp tools rather than chopping, though this still needs validation.\n- **Demographics:** Subadults seem underrepresented. The only young child had its head.\n- **Context:** The authors say head removal is a widespread phenomenon, not a south-west Asian tradition brought by early farmers.\n\nFull osteological analysis is still pending.",
     "verdict": "partial",
     "facts_found": 3,
     "facts_total": 5,
     "judge_note": "Gets Vráble, the at least 77 headless skeletons in the ditch and the pending analyses, but omits the interpretations (headhunting, massacre, sacrifice, burial rite) and the Herxheim/Asparn-Schletz/Talheim comparisons.",
     "wall_seconds": 7.45,
     "retrieval_seconds": 2.15,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6351,
     "cache_read_tokens": 0,
     "output_tokens": 831,
     "cost_usd_list": 0.033718,
     "retrieved_chunks": [
      "W7163143446#13",
      "W7163143446#6",
      "W7163143446#11",
      "W7163143446#8",
      "W7163143446#23",
      "W7163143446#10",
      "W7163143446#7",
      "W7163143446#14"
     ]
    },
    "okf": {
     "answer": "At Vráble, a Late LBK enclosed settlement in south-west Slovakia, excavators found a mass deposition of headless skeletons at the bottom of the outer ditch, east of entrance 2 (Furholt et al. 2026).\n\n- **Numbers:** The first find was in 2022, when there were 37 skeletons, 36 of them headless. Excavation through 2024 brought the total to at least 77 headless skeletons and one complete skeleton within the first 25 m. The end of the deposition has not been found. The other ditch terminal, 158 m away, held at least two more headless skeletons.\n- **Placement:** The bodies lie in varied positions, with no regular orientation. Most are well articulated and sometimes overlie each other. There are no grave goods. The only young child is the only one still with its head.\n- **Date:** Radiocarbon dates fall at 5300/5210–5050/4950 cal BCE. The deposition was probably brief, \"a few years, maybe even less\". The calibration plateau means the dates can't show whether it was a one-off event or a recurring practice.\n- **Missing heads:** Only a few skull and mandible fragments were found, so the heads' fate is unknown. The authors consider several explanations:\n  - headhunting;\n  - massacre, though the patterned placement of other headless bodies argues against it;\n  - a longer-term practice with motives other than violence;\n  - human sacrifice;\n  - a burial rite focused on the skull.\n\n  One idea is that skulls were displayed in houses.",
     "verdict": "partial",
     "facts_found": 3,
     "facts_total": 5,
     "judge_note": "The answer correctly covers Vráble, the at least 77 headless skeletons in the ditch, the missing heads and the open interpretations, but omits the comparison with Herxheim, Asparn-Schletz and Talheim and the fact that analyses (aDNA, isotopes, kinship) are still under way.",
     "wall_seconds": 11.73,
     "retrieval_seconds": null,
     "turns": 7,
     "input_tokens": 8,
     "cache_creation_tokens": 6223,
     "cache_read_tokens": 16506,
     "output_tokens": 1259,
     "cost_usd_list": 0.039148600000000006,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "The find comes from the LBK settlement of Vráble-Veľké Lehemby in south-west Slovakia (Furholt et al. 2026). It was uncovered in 2022 at the bottom of the 1.3 km double ditch enclosing one of the site's three neighbourhoods.\n\n**Scale:** The count rose from 37 skeletons (36 headless) in 2022 to at least 77 headless skeletons plus one complete one by 2024. All lay within the first 25 m of the ditch terminal east of entrance 2, and the end of the deposit has not been reached. Across the site, 112 individuals have been identified so far.\n\n**Date:** Radiocarbon dates give about 5300/5210–5050/4950 cal BCE. The bodies were probably placed over a few years at most, though hours or days are possible. A flat stretch of the calibration curve prevents anything more precise.\n\n**Observations:**\n- The bodies were not far decomposed when deposited.\n- Cutmarks on the neck vertebrae suggest the heads were removed with sharp tools.\n- Mandibles are almost absent, so whole heads were apparently taken. The heads themselves have not been found.\n- There are no grave goods, but pebbles lie among the bones.\n- Headless \"pairs\" also occur at each of the four southern entrances, so the deposit is part of a wider pattern.\n\n**Interpretation:** This is unresolved. The authors consider headhunting, massacre, sacrifice or a special burial rite. They note the patterned placements argue against a single massacre. They suggest tensions within the settlement community, since the enclosure and its entrances face away from the other neighbourhoods. They link the find to other late-LBK ditch depositions such as Asparn-Schletz and Herxheim.",
     "verdict": "partial",
     "facts_found": 4,
     "facts_total": 5,
     "judge_note": "The answer correctly covers Vráble, the 77+ headless skeletons in the ditch, the missing heads, the open interpretations and the Herxheim/Asparn-Schletz comparison, but it omits Talheim and the ongoing aDNA, isotope and kinship analyses.",
     "wall_seconds": 9.16,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 559,
     "cache_read_tokens": 158219,
     "output_tokens": 986,
     "cost_usd_list": 0.0279219,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q16",
   "type": "synthesis",
   "question": "How well established is the absolute chronology of the end of the LBK across these papers, and what methodological problems do they report?",
   "ground_truth": "Not well established. Furholt et al.: LBK chronologies are largely based on pottery decoration and are not controlled by absolute dates because of the calibration plateau (5210-5000 cal BCE). Kiosak & Kotova: the Kyiv laboratory series from 1998-2008 is suspicious and fails cross-laboratory validation, dates on total organic carbon in pottery (TOCC) are unreliable, and the end of LBK falls on the plateau; they use Bayesian modelling in OxCal. Kroon et al.: the dataset is sparse (482 dates reviewed, a smaller high-quality subset) and mostly charred cereals with a risk of residual or intrusive dates; the overarching typochronology model has poor agreement (Amodel 7.4%) while regional models fit well; they present two alternative models and call for renewed absolute dating.",
   "key_facts": [
    "calibration plateau",
    "Kyiv lab series 1998-2008 and TOCC dates unreliable",
    "sparse dataset, charred cereals risk residual or intrusive",
    "typochronology vs radiocarbon mismatch, two models",
    "call for more absolute dating"
   ],
   "approaches": {
    "none": {
     "answer": "I can't see any papers in this conversation, so I can't say what these particular authors report. Here is what I know about the topic in general.\n\n**How well established:** The end of the LBK is only moderately well dated. It is usually placed around 5000-4900 cal BC, and Bayesian models for the Alsace/Rhine area (e.g. Denaire et al. 2017) put the transition to Hinkelstein/Grossgartach/Villeneuve-Saint-Germain at about 4950 cal BC. Regional end dates vary by a century or more, and the collapse seems diachronic rather than simultaneous.\n\n**Commonly reported problems:**\n- **Calibration plateaus and wiggles** in the curve around the late 6th millennium widen the probability ranges.\n- **Few reliable dates** for the final phases, with many old, low-precision conventional dates.\n- **Sample quality:** old-wood effects on charcoal, uncertain context or association, and bone versus charcoal offsets.\n- **Dependence on typological phasing:** Bayesian models need secure stratigraphy or seriation, and ceramic phase boundaries are fuzzy.\n- **Poor stratigraphy** at many sites, with sparse late-LBK settlements and uneven regional coverage.\n- **Terminal-LBK events** such as Talheim, Schletz and Herxheim are hard to date precisely or to link to a single episode.\n\nIf you share the papers or their titles, I can compare them directly.",
     "verdict": "partial",
     "facts_found": 1,
     "facts_total": 5,
     "judge_note": "The answer gives general background rather than the papers' findings: it matches the calibration plateau and vaguely touches on sparse dates and typology, but misses the Kyiv lab/TOCC problems, the charred-cereal residuality risk, the two alternative models with poor Amodel, and the call for more dating.",
     "wall_seconds": 10.87,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 672,
     "cache_read_tokens": 0,
     "output_tokens": 948,
     "cost_usd_list": 0.012171999999999999,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The end of the LBK is **not well established** in absolute terms; the papers describe it as largely typochronological and only loosely constrained by radiocarbon.\n\n**Evidence and estimates**\n- Kroon et al. model the end of LBK occupation: Limburg c. 5107–5013 BC (68.3%); Rhineland 5186–4871 BC (68.3%), with an end boundary lasting up to 212 years. For Hesbaye, the earliest appearance of the post-LBK Blicquy/Villeneuve-Saint-Germain (BVSG) is 5377–4879 BC (68.3%). The BVSG overlap with LBK therefore ranges from a 106-year gap to a 415-year overlap.\n- Kiosak & Kotova argue the LBK in the Ukrainian region did not outlast its Central European core.\n\n**Methodological problems**\n- Current chronologies rest mainly on pottery decoration, not absolute dates.\n- Sparse radiocarbon datasets: the Rhineland post-LBK model has only 2 dates and low convergence, and there are no Grossgartach dates.\n- Calibration curve plateaus hinder dating (Vráble).\n- Summed-probability methods suffer from sample bias and curve-shape artefacts.\n- Different assumptions (typochronology vs. radiocarbon priority) give different models (Models 6 and 7).\n- Cross-laboratory inconsistencies and unreliable dates (e.g., pottery organic content) occur.",
     "verdict": "partial",
     "facts_found": 3,
     "facts_total": 5,
     "judge_note": "The answer covers the calibration plateau, the typochronology-versus-radiocarbon divergence with two models, and sparse data with cross-lab and pottery-date problems only vaguely; it omits the Kyiv 1998-2008 series, the residual/intrusive cereal-date risk and the call for renewed dating.",
     "wall_seconds": 5.26,
     "retrieval_seconds": 1.96,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6625,
     "cache_read_tokens": 0,
     "output_tokens": 496,
     "cost_usd_list": 0.031464,
     "retrieved_chunks": [
      "W7163143446#17",
      "W7225183261#25",
      "W7225183261#35",
      "W7225183261#12",
      "W7225183261#30",
      "W7225183261#4",
      "W7204009876#5",
      "W7225183261#33"
     ]
    },
    "okf": {
     "answer": "The papers only partly establish an absolute date for the end of the LBK. The end falls on the late 6th-millennium plateau in the calibration curve, so the dates are broad, and they differ by region.\n\n**Modelled end dates**\n- **Buh–Dniester:** the LBK ends 5211–5010 cal BC (95.4%), or 5207–5048 (68.3%).\n- **Limburg:** latest LBK is 5107–5013 BC (68.3%). The end boundary lasts 0–38 years at 68.3% and 0–109 years at 95.4%.\n- **Hesbaye:** latest LBK is 5024–4952 BC (68.3%). The end boundary lasts 0–42 and 0–110 years.\n- **Rhineland:** latest LBK is 5186–4871 BC (68.3%). The end boundary lasts 0–212 and 0–393 years.\n- **Vráble:** the mass deposition falls anywhere between 5210 and about 4950 cal BCE.\n\n**Methodological problems reported**\n- **Plateau:** the flat calibration curve at c. 5210–5000 cal BC stretches the calibrated ranges. It blocks finer dating at Vráble and in Buh–Dniester, and it obscures whether the deposition was a single event or a recurring practice.\n- **Pottery-based chronologies:** the authors say the idea of a \"final LBK phase\" is not firmly established by absolute dates.\n- **Sparse, risky dates (Rhineland–Hesbaye):** only 62 LBK dates survive from 482 reviewed. Articulated bone is missing because bone preserves poorly in decalcified loess, and the risk of residual or intrusive dates remains high. Other dates are only usable as terminus post quem because of old-wood effects.\n- **Post-LBK overlap:** the Rhineland has just 2 post-LBK dates, so no conclusion on a gap is possible. In Hesbaye the relation to BVSG ranges from a 106-year gap to a 415-year overlap (68.3%).",
     "verdict": "partial",
     "facts_found": 2,
     "facts_total": 5,
     "judge_note": "The answer covers the calibration plateau and the sparse dataset with residual/intrusive risk, but omits the Kyiv lab/TOCC problems, the poor-fitting typochronology model with two alternative models, and the call for renewed absolute dating.",
     "wall_seconds": 16.42,
     "retrieval_seconds": null,
     "turns": 10,
     "input_tokens": 8,
     "cache_creation_tokens": 9594,
     "cache_read_tokens": 20438,
     "output_tokens": 1967,
     "cost_usd_list": 0.0601058,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "**Short answer:** the end of the LBK is only loosely constrained. It falls around 5100-4950 BC in most areas, but the uncertainty is often 100+ years and the data are sparse.\n\n**Estimated end dates**\n- **Limburg:** 5107-5013 BC (68%). **Hesbaye:** 5024-4952 BC. **Rhineland:** 5186-4871 BC, with an end boundary of up to 212 years (Kroon et al.).\n- **Ukraine/Moldova:** 5207-5048 cal BC (68%) (Kiosak & Kotova).\n- **Vráble:** the site ends c. 5000-4950 BC (Furholt et al.).\n- **Alsace:** the LBK ends c. 4950 BC, and Hinkelstein arrives 40-175 years later according to Denaire et al. 2017, though Waldvogel notes the two may overlap.\n\n**Methodological problems reported**\n- **Calibration plateau, c. 5210-5000 BC:** this blocks finer resolution (Furholt, Kiosak).\n- **Sparse or poor dates:** there are only five usable dates for the Rhineland and none for Großgartach, which prevents conclusions on a gap. Dates on charred seeds carry a residuality or intrusion risk, and no articulated bone dates exist.\n- **Typochronology versus radiocarbon:** the overarching model fits poorly (Amodel 7.4%). Resolving this requires either heavy outlier-cleaning, which shortens the LBK to 58-176 years, or overlapping chronozones.\n- **Key dates fail as outliers:** Maastricht-Cannerberg is an outlier, and Uckendorf is a 100% early outlier.\n- **Unreliable dates:** pottery TOCC dates and the Kyiv 1998-2008 series are unreliable (Kiosak & Kotova).",
     "verdict": "partial",
     "facts_found": 4,
     "facts_total": 5,
     "judge_note": "The answer covers the calibration plateau, the unreliable Kyiv and TOCC dates, the sparse charred-seed data and the typochronology mismatch with two alternative models, but it omits the call for renewed absolute dating.",
     "wall_seconds": 10.24,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 572,
     "cache_read_tokens": 158219,
     "output_tokens": 1180,
     "cost_usd_list": 0.0299139,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q17",
   "type": "synthesis",
   "question": "What do the papers say about whether archaeological cultures defined by pottery style correspond to real social groups?",
   "ground_truth": "Waldvogel (Alsace): funerary variability (body orientation and position, grave layout, grave goods) frequently transcends chrono-cultural divisions based on ceramic styles and cannot be correlated with sex or age, so it questions the view of archaeological cultures as coherent social entities and suggests social units rooted in territories. Kiosak & Kotova: the 'Buh-Dniester culture' is effectively an umbrella term for diverse sites, Savran-style pottery is not exclusive to one period, and ceramic styles of hunter-gatherer pottery do not appear to have purely chronological significance; isolated traits attributed to cultures can mislead.",
   "key_facts": [
    "Alsace: funerary variability transcends ceramic-defined cultures",
    "Alsace: not correlated with sex or age",
    "Buh-Dniester is an umbrella term",
    "Savran pottery not exclusive to the interval, styles not purely chronological"
   ],
   "approaches": {
    "none": {
     "answer": "I don't have any specific papers in front of me, so this is a summary of the broader literature as I know it.\n\n**Mixed verdict: pottery cultures sometimes track social groups, but the match is not automatic.**\n\n- **Culture-historical view** (Kossinna, early Childe): treated pottery styles as the signature of peoples. Later archaeologists largely rejected this.\n- **Ethnoarchaeology** (Hodder's 1982 Baringo study, Wiessner 1983, Stark on the Kalinga): style boundaries often cross ethnic lines, or only mark groups in certain contexts and for certain artifact types. Technology and social learning networks can matter more than identity.\n- **Transmission models** (Shennan, Eerkens & Lipo): style patterns can arise from neutral copying and drift, without bounded \"peoples.\"\n- **Ancient DNA**: results are mixed. Corded Ware and Yamnaya show strong genetic links, and Bell Beaker spread by migration into Britain (Olalde et al. 2018, with roughly 90% turnover). In Iberia, however, the Beaker package spread largely without wholesale population replacement. Critics such as Furholt warn against equating cultures with populations.\n\nOverall, treat archaeological cultures as heuristic categories, not as given social units.",
     "verdict": "wrong",
     "facts_found": 0,
     "facts_total": 4,
     "judge_note": "The answer admits it has no access to the papers and gives a general literature overview, missing every paper-specific fact (Alsace funerary variability, Buh-Dniester umbrella term, Savran pottery).",
     "wall_seconds": 11.01,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 669,
     "cache_read_tokens": 0,
     "output_tokens": 901,
     "cost_usd_list": 0.01169,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The excerpts (all from Waldvogel's 2026 study of the Alsace Plain, 5300–4400 BC) argue that pottery-based cultures do **not** correspond to coherent, homogeneous social groups.\n\n- **Funerary variability crosses cultural lines.** Patterns in body orientation and position, grave layout and grave goods occur independently of the ceramic-defined divisions (LBK, Hinkelstein, Grossgartach, Rössen). They are also unrelated to biological sex or age at death.\n- **Likely social units at a different scale.** Recurrent gestures within specific areas (three or four \"facies\" and several \"sub-facies\") may reflect social units rooted in particular territories, such as clans or lineages. These units shared a wider Danubian identity.\n- **Shared material culture doesn't mean one group.** Groups could distinguish themselves through subtle variations while sharing pottery styles. The study cites Hodder, Renfrew, and Roberts & Vander Linden.\n- **Ceramic change ≠ population change.** Style shifts may reflect local transformations or interaction. Example: extended supine burial in Lower Alsace resembles Hinkelstein practice in the Worms region.\n\nThe author notes that facies are statistical tendencies, and that their exact social significance remains hard to determine.",
     "verdict": "partial",
     "facts_found": 2,
     "facts_total": 4,
     "judge_note": "Covers both Alsace facts but misses the Kiosak & Kotova points on the Buh-Dniester umbrella term and Savran pottery styles.",
     "wall_seconds": 5.24,
     "retrieval_seconds": 1.78,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6292,
     "cache_read_tokens": 0,
     "output_tokens": 433,
     "cost_usd_list": 0.029502,
     "retrieved_chunks": [
      "W7202185506#8",
      "W7202185506#31",
      "W7202185506#0",
      "W7202185506#9",
      "W7202185506#22",
      "W7202185506#7",
      "W7202185506#33",
      "W7202185506#4"
     ]
    },
    "okf": {
     "answer": "The bundle's papers say pottery-defined cultures are probably not coherent social groups. The evidence is indirect, though, and mostly comes from two of the five papers.\n\n- **Waldvogel 2026 (Alsace Plain):** This paper gives the most direct answer. Funerary variability is structured, but it isn't tied to sex or age. It also cuts across ceramic-defined cultures (LBK, Hinkelstein, Grossgartach, Rössen). In the LBK, Lower Alsace burials mostly have the head west (31%) and Upper Alsace burials mostly have the head east (92%). House plans, animal economy and ceramic styles also differ between the two areas. The paper reads these as territorially rooted social units, such as clans or lineages. They may share a wider Danubian identity. Practices reproduced over generations may persist across ceramic-defined cultures. The paper says its results \"question archaeological cultures (defined by pottery) as coherent social entities\", citing Hodder, Renfrew, and Roberts & Vander Linden. It calls the case most convincing for the LBK and only suggestive for the other cultures.\n- **Kroon et al. 2026 (LBK chronozones):** Ceramic styles are longer-lived and partly contemporaneous. Chronozones overlap, so stylistic sequences are not clean chronological or social units.\n- **Kiosak & Kotova 2026 (Southern Buh–Dniester):** The Savran phase and Buh–Dniester culture are described as questionable. Even so, Savran-style pottery turns up at post-LBK sites.",
     "verdict": "partial",
     "facts_found": 3,
     "facts_total": 4,
     "judge_note": "The Alsace points are fully covered, but the Kiosak & Kotova part is weak: it calls Buh-Dniester 'questionable' rather than an umbrella term for diverse sites, only implies Savran pottery is not exclusive, and never says the styles are not purely chronological.",
     "wall_seconds": 12.08,
     "retrieval_seconds": null,
     "turns": 7,
     "input_tokens": 6,
     "cache_creation_tokens": 6322,
     "cache_read_tokens": 10255,
     "output_tokens": 1352,
     "cost_usd_list": 0.0398455,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "Three of the papers address this directly. All doubt that pottery-defined cultures are coherent social units.\n\n- **Waldvogel (Alsace):** Funerary \"facies\" and \"sub-facies\" recur in specific territories and cut across the ceramic-based LBK, Hinkelstein, Grossgartach and Rössen divisions. They do not correlate with sex or age. She argues this challenges the view of archaeological cultures as coherent, homogeneous social entities. At Obernai, DNA shows biological continuity between Grossgartach and Rössen individuals despite their different ceramics. Changes in ceramic style may therefore reflect transformations within local groups rather than the arrival of new populations.\n- **Kroon et al. (Rhineland–Hesbaye):** Ceramic styles once treated as chronological markers are longer-lived and partly contemporaneous. Their radiocarbon models suggest that stylistic traits appear and persist at different times in different regions.\n- **Kiosak & Kotova (Ukraine):** They question whether the Buh–Dniester culture and its Savran phase really exist, calling the label an umbrella term. Pottery styles lack purely chronological meaning, and attributing isolated traits to specific cultures can be misleading.\n\nFurholt et al. only note that LBK chronologies built on pottery decoration are not yet well established by absolute dating.",
     "verdict": "correct",
     "facts_found": 4,
     "facts_total": 4,
     "judge_note": "All four key facts are present, though the Savran point is phrased as doubting the phase rather than saying the pottery is not exclusive to one period.",
     "wall_seconds": 8.11,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 569,
     "cache_read_tokens": 158219,
     "output_tokens": 835,
     "cost_usd_list": 0.0264519,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q18",
   "type": "unanswerable",
   "question": "What was the average stature of the adults buried at Vráble, and what ratio of males to females did the osteological analysis find?",
   "ground_truth": "Not reported. The Vráble paper presents only preliminary osteological observations (for example that subadults seem underrepresented) and states the full osteological analysis is still to be carried out.",
   "key_facts": [
    "not stated in the dataset"
   ],
   "approaches": {
    "none": {
     "answer": "I don't know the average adult stature or the male-to-female ratio from the osteological analysis of the Vráble burials. I'm not confident of the figures, and I'd rather not guess at them.\n\nI do know that Vráble in southwestern Slovakia is a large Linearbandkeramik (LBK) settlement, c. 5250-4950 BCE, studied by the Kiel and Nitra teams. Its burials have also been used in ancient DNA and kinship studies. The specific stature and sex-ratio results would be in the osteological sections of the project's publications, so I'd check those directly.\n\nIf you can share the paper or report, I can help interpret the numbers.",
     "verdict": "partial",
     "facts_found": 0,
     "facts_total": 1,
     "judge_note": "The answer avoids inventing figures, but it frames the gap as its own ignorance and implies the results exist in publications, when the paper says the osteological analysis is still to be done.",
     "wall_seconds": 5.6,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 674,
     "cache_read_tokens": 0,
     "output_tokens": 408,
     "cost_usd_list": 0.00678,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The excerpts don't give either figure. They contain no average stature for the adults buried at Vráble and no male-to-female ratio from the osteological analysis.\n\nThe authors say the full osteological analysis is still to be carried out and that they offer only preliminary observations. Those observations are:\n\n- Subadult skeletons seem to be underrepresented compared to adults, judging by epiphyseal closure.\n- The only skeleton of a young child is also the only individual in the mass deposition that still had its head.\n- One adult male was buried at the centre of entrance 1 with six ceramic containers and a long flint blade. This is a single burial, not a sex ratio.\n\nThe excerpts also say the 2016–2017 grave remains are published in Hukeľová & Krošláková 2021 and Müller-Scheeßel & Hukeľová 2020. Those publications show a broad demographic structure and discuss pathologies and survived injuries. Stature and sex ratios may be in them, but the excerpts don't say.",
     "verdict": "correct",
     "facts_found": 1,
     "facts_total": 1,
     "judge_note": "The answer says plainly that neither stature nor sex ratio is reported, matches the preliminary-observation framing, and invents no figures.",
     "wall_seconds": 4.15,
     "retrieval_seconds": 2.08,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6787,
     "cache_read_tokens": 0,
     "output_tokens": 322,
     "cost_usd_list": 0.030372,
     "retrieved_chunks": [
      "W7163143446#14",
      "W7163143446#15",
      "W7163143446#6",
      "W7163143446#1",
      "W7163143446#0",
      "W7163143446#13",
      "W7163143446#24",
      "W7163143446#2"
     ]
    },
    "okf": {
     "answer": "The bundle doesn't give either figure, so I can't answer from it.\n\nThe Vráble files I read cover the mass deposition of headless skeletons, the taphonomic observations, the regular burials and the dating. None of them reports an average adult stature or a male-to-female ratio. A search of the whole bundle for stature, ratio, average, females and similar terms also found nothing relevant to Vráble.\n\nThe bundle describes the osteology only as preliminary, made on site in 2021–2024. It gives these related facts:\n- There are at least 77 headless skeletons and one complete skeleton in the ditch.\n- Subadults appear underrepresented.\n- Burial G40 is a pregnant woman.\n\nIt also says that sex and age, genetics and isotope analyses are still planned. The full osteological report may be published elsewhere.",
     "verdict": "correct",
     "facts_found": 1,
     "facts_total": 1,
     "judge_note": "The answer says neither stature nor sex ratio is reported, notes the osteology is preliminary, and invents no figures.",
     "wall_seconds": 14.95,
     "retrieval_seconds": null,
     "turns": 8,
     "input_tokens": 14,
     "cache_creation_tokens": 8463,
     "cache_read_tokens": 37657,
     "output_tokens": 1293,
     "cost_usd_list": 0.0505757,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "The papers don't give either figure. The Vráble paper (Furholt et al. 2026) says the full osteological analysis is still pending, so no average stature or male-to-female ratio is reported.\n\nWhat it does report is preliminary:\n- At least 112 individuals have been identified (as of summer 2025), and the count will probably rise.\n- The mass deposition includes at least 77 headless skeletons and one complete skeleton.\n- Judging by epiphyseal closure, subadults seem underrepresented relative to adults.\n- The only young child in the mass deposition is also the only individual that still had a head.\n- A few individuals are sexed individually, for example adult males in the entrance 1 burials and a pregnant woman buried at the edge of the ditch. No overall sex ratio is given.",
     "verdict": "correct",
     "facts_found": 1,
     "facts_total": 1,
     "judge_note": "The answer says neither stature nor sex ratio is reported because the full osteological analysis is pending, and its extra preliminary details do not invent the requested figures.",
     "wall_seconds": 4.64,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 574,
     "cache_read_tokens": 158219,
     "output_tokens": 367,
     "cost_usd_list": 0.0217919,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q19",
   "type": "unanswerable",
   "question": "Which mitochondrial haplogroups did the ancient DNA analysis of the Vráble headless skeletons reveal?",
   "ground_truth": "Not reported. aDNA analyses of the Vráble material are listed as under way; aDNA results are mentioned only for Asparn-Schletz (92 individuals).",
   "key_facts": [
    "not stated in the dataset"
   ],
   "approaches": {
    "none": {
     "answer": "I don't know which mitochondrial haplogroups were reported for the headless skeletons from Vráble, so I won't guess. I know Vráble as a large Linear Pottery culture (LBK) settlement in southwestern Slovakia, dated to roughly 5250-4950 BCE. It has been the subject of ancient DNA work on kinship and social organisation. But I can't reliably link specific haplogroups to the headless burials. The original publication or its supplementary data would list the haplogroup assignments per individual.",
     "verdict": "correct",
     "facts_found": 1,
     "facts_total": 1,
     "judge_note": "The answer declines to name any haplogroups and invents none, though it frames this as its own uncertainty and implies a publication lists them, while the ground truth says the Vráble aDNA analyses are still under way.",
     "wall_seconds": 5.52,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 677,
     "cache_read_tokens": 0,
     "output_tokens": 395,
     "cost_usd_list": 0.0066619999999999995,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The excerpts don't say. They report no mitochondrial haplogroup results for the Vráble headless skeletons. The paper states that aDNA, eDNA, stable isotopes, bioerosion and biomarker analyses are still under way, and that questions such as the biological and social kinship of the headless skeletons to the Vráble inhabitants are still being pursued.\n\nThe only aDNA result mentioned comes from a different site, Asparn-Schletz in Lower Austria. There, data from 92 individuals suggest that those killed likely did not represent the inhabitants of the settlement. The excerpt cuts off there and gives no haplogroups.",
     "verdict": "correct",
     "facts_found": 1,
     "facts_total": 1,
     "judge_note": "The answer says no haplogroups are reported, notes that the aDNA analyses are still under way, and attributes the only aDNA result to Asparn-Schletz without inventing anything.",
     "wall_seconds": 3.37,
     "retrieval_seconds": 1.82,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 6993,
     "cache_read_tokens": 0,
     "output_tokens": 210,
     "cost_usd_list": 0.030076000000000002,
     "retrieved_chunks": [
      "W7163143446#34",
      "W7163143446#24",
      "W7163143446#6",
      "W7163143446#1",
      "W7163143446#0",
      "W7163143446#23",
      "W7163143446#21",
      "W7163143446#35"
     ]
    },
    "okf": {
     "answer": "The bundle doesn't say. It states that aDNA analyses of the Vráble headless skeletons are \"under way and not reported\" (Furholt et al. 2026, `papers/W7163143446.md`). Genetics is listed only as a planned analysis (`findings/vrable-taphonomic-observations.md`). I found no mitochondrial haplogroup results anywhere in it.",
     "verdict": "correct",
     "facts_found": 1,
     "facts_total": 1,
     "judge_note": "The answer states that no haplogroup results are reported and that the aDNA work is still under way, and it invents no results.",
     "wall_seconds": 6.85,
     "retrieval_seconds": null,
     "turns": 4,
     "input_tokens": 8,
     "cache_creation_tokens": 5349,
     "cache_read_tokens": 20259,
     "output_tokens": 494,
     "cost_usd_list": 0.028377900000000005,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "The paper doesn't report any mitochondrial haplogroups for the Vráble headless skeletons. Furholt et al. (2026) say that aDNA, eDNA, stable isotope, bioerosion and biomarker analyses are still under way, and they present only the archaeological context, new radiocarbon dates and preliminary osteological and taphonomic observations. They expect population genetics and genetic kinship analyses to help clarify the biological and social relationships of the individuals, but no results are given.\n\nThe only aDNA result in the paper concerns a different site. At Asparn-Schletz (Austria), data from 92 individuals suggest the people killed there were probably not the settlement's inhabitants, because their level of biological relatedness was too low (Gelabert et al. 2025). No haplogroups are mentioned for that site either.",
     "verdict": "correct",
     "facts_found": 1,
     "facts_total": 1,
     "judge_note": "The answer says plainly that no haplogroups are reported, notes that aDNA work is still under way, and correctly limits the aDNA results to Asparn-Schletz without inventing anything.",
     "wall_seconds": 4.29,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 577,
     "cache_read_tokens": 158219,
     "output_tokens": 302,
     "cost_usd_list": 0.0211539,
     "retrieved_chunks": null
    }
   }
  },
  {
   "id": "Q20",
   "type": "unanswerable",
   "question": "What radiocarbon dates were obtained for features at Zwięczyca site 3?",
   "ground_truth": "None reported. The radiocarbon dating in the Polish paper covers eight samples from Łańcut only; Zwięczyca is dated by typological features (Pre-Music Note to Želiezovce phases), not by radiocarbon dates in this paper.",
   "key_facts": [
    "not stated for Zwięczyca",
    "the eight dates are from Łańcut"
   ],
   "approaches": {
    "none": {
     "answer": "I don't have reliable knowledge of the specific radiocarbon dates from Zwięczyca site 3 (near Rzeszów, southeastern Poland), so I can't give you the numbers without guessing.\n\nWhat I can say with some confidence is general context: Zwięczyca is a multi-period archaeological locality in the Rzeszów area of the Podkarpacie region, and sites there have been associated with Neolithic and later prehistoric occupation. But I can't confidently attribute specific uncalibrated BP values, calibrated ranges, lab codes, or feature numbers to site 3 specifically.\n\nFor accurate figures, I'd suggest:\n\n- The original excavation report or publication (likely in *Sprawozdania Archeologiczne*, *Materiały Archeologiczne*, or a Rzeszów-area regional journal)\n- The Context of Radiocarbon Dates database or the Radiocarbon journal date lists, which record lab numbers (e.g., Gd-, Poz-, Ki-) with site names\n- The Institute of Archaeology at the University of Rzeszów, which has worked in the area\n\nIf you can share the paper or a passage, I can help interpret or summarize the dates.",
     "verdict": "correct",
     "facts_found": 1,
     "facts_total": 2,
     "judge_note": "The answer declines to give Zwięczyca dates and invents none, but it frames this as its own lack of knowledge rather than saying none were reported, and it omits that the eight dates are from Łańcut.",
     "wall_seconds": 5.4,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 664,
     "cache_read_tokens": 0,
     "output_tokens": 392,
     "cost_usd_list": 0.006579999999999999,
     "retrieved_chunks": null
    },
    "rag": {
     "answer": "The excerpts don't give any radiocarbon dates for Zwięczyca site 3. They report AMS dating only for Łańcut site 3: eight samples (seven cereal grains plus one wheat spikelet fork) from features there, dated at the Poznań Radiocarbon Laboratory. The Łańcut dates visible in Table 1 include:\n\n- Poz-193691: 5,395 ± 35 BP (4340–4063 cal BC), *Triticum aestivum*, feature 102\n- Poz-193692: 5,330 ± 40 BP (4323–4047 cal BC), *T. dicoccon*, feature 102\n- Poz-192621: 6,180 ± 40 BP (5286–5003 cal BC), *T. dicoccon*, feature 93\n- Poz-192691: 6,140 ± 40 BP (5213–4954 cal BC), *T. dicoccon*, feature 23\n\nThe table is cut off after that, and the excerpts say nothing about Zwięczyca having been radiocarbon dated. The methods state that samples were selected from features at Łańcut only.",
     "verdict": "correct",
     "facts_found": 2,
     "facts_total": 2,
     "judge_note": "The answer says no radiocarbon dates are reported for Zwięczyca and attributes the eight AMS samples to Łańcut; the extra Łańcut details don't contradict the ground truth.",
     "wall_seconds": 4.39,
     "retrieval_seconds": 1.85,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 7469,
     "cache_read_tokens": 0,
     "output_tokens": 365,
     "cost_usd_list": 0.03353,
     "retrieved_chunks": [
      "W7204009876#5",
      "W7204009876#15",
      "W7166088190#6",
      "W7204009876#6",
      "W7204009876#12",
      "W7166088190#5",
      "W7204009876#3",
      "W7166088190#10"
     ]
    },
    "okf": {
     "answer": "The bundle doesn't contain any radiocarbon dates for Zwięczyca site 3.\n\n- The Zwięczyca site page, \"Rzeszów-Zwięczyca site 3\", has no dates. It gives only a typological placement of the features. They span the Pre-Music Note phase to the Želiezovce phase of the LBK, and the recent features \"likely represent the younger phase\".\n- The source paper (Juźwińska et al. 2026) reports eight AMS dates, but all eight are from **Łańcut site 3** (the paper's other site), not Zwięczyca.\n- The Łańcut dates include emmer at about 5286–4846 cal BC (2σ). Feature 102 gave bread wheat at 4340–4063 cal BC and emmer at 4323–4047 cal BC.\n\nIf you meant the Łańcut dates, I can list them from the radiocarbon dating page.",
     "verdict": "correct",
     "facts_found": 2,
     "facts_total": 2,
     "judge_note": "The answer says no radiocarbon dates are reported for Zwięczyca and attributes all eight dates to Łańcut; the extra Łańcut details do not contradict the ground truth.",
     "wall_seconds": 8.49,
     "retrieval_seconds": null,
     "turns": 4,
     "input_tokens": 8,
     "cache_creation_tokens": 14099,
     "cache_read_tokens": 21511,
     "output_tokens": 688,
     "cost_usd_list": 0.0654431,
     "retrieved_chunks": null
    },
    "full": {
     "answer": "The paper doesn't report any radiocarbon dates for features at Zwięczyca site 3. Radiocarbon dating (AMS, Poznań Radiocarbon Laboratory) was done only on eight samples from features at Łańcut site 3. Zwięczyca's LBK attribution rests on pottery: its multiphase settlement runs from the earliest Pre-Music Note phase to the late Želiezovce phase, with at least 47 longhouses (Dębiec 2014).\n\nFor comparison, the Łańcut results were:\n- **Six emmer grains and one Triticum sp. spikelet fork** gave LBK-range ages, from about 6,080 ± 40 to 6,190 ± 40 BP. The calibrated 2σ ranges span 5294–4846 cal BC.\n- **Feature 102 (emmer and bread wheat)** gave about 5,330–5,395 BP (4340–4047 cal BC), which points to the later Malice culture.",
     "verdict": "correct",
     "facts_found": 2,
     "facts_total": 2,
     "judge_note": "The answer says the paper gives no radiocarbon dates for Zwięczyca, dates it by typology instead, and correctly attributes the eight dates to Łańcut.",
     "wall_seconds": 5.35,
     "retrieval_seconds": null,
     "turns": 1,
     "input_tokens": 2,
     "cache_creation_tokens": 564,
     "cache_read_tokens": 158219,
     "output_tokens": 396,
     "cost_usd_list": 0.0220419,
     "retrieved_chunks": null
    }
   }
  }
 ]
}
