{
  "corpus": "contradictions",
  "results": [
    {
      "id": "flu-1918-deaths",
      "question": "how many people died in the 1918 influenza pandemic?",
      "tags": [
        "disputed",
        "epidemiology",
        "numeric"
      ],
      "session_id": "s_06FZA17P8XNM1NHD",
      "status": "done",
      "claims": 36,
      "scorecard": {
        "session_id": "s_06FZA17P8XNM1NHD",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0110 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "47 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "36 of 36 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 22.22222222222222,
            "unit": "%",
            "detail": "6 distinct source(s); the largest contributed 8 of 36 claims (https://archive.cdc.gov/www_cdc_gov/flu/pandemic-resources/1918-commemoration/1918-pandemic-history.htm)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.000307,
            "unit": "usd",
            "detail": "$0.0110 over 36 claims = $0.0003 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 47,
            "unit": "calls",
            "detail": "llm 46 · search 1"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "36 of 36 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 33.333333333333336,
            "unit": "%",
            "detail": "36 claim(s) render as 24 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "value": 16.666666666666668,
            "unit": "%",
            "detail": "4 contradiction(s) involving 6 of 36 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "measured",
            "unit": "edges",
            "detail": "0 of 4 conflict(s) resolved as staleness by publication date — check whether the sources carried publication dates at all"
          },
          {
            "name": "grounding rate",
            "status": "blocked",
            "reason": "no claim was re-read: §11.5.2's grounding pass needs escrow left over and a fetcher configured"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    },
    {
      "id": "population-peak",
      "question": "when will the world population peak and at what level?",
      "tags": [
        "disputed",
        "demography",
        "numeric"
      ],
      "session_id": "s_06FZA1N1KDMQ4MCT",
      "status": "done",
      "claims": 63,
      "scorecard": {
        "session_id": "s_06FZA1N1KDMQ4MCT",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0214 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "88 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "63 of 63 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 12.698412698412698,
            "unit": "%",
            "detail": "12 distinct source(s); the largest contributed 8 of 63 claims (https://populationmatters.org/news/2024/04/the-world-of-population-projections)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.000339,
            "unit": "usd",
            "detail": "$0.0214 over 63 claims = $0.0003 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 88,
            "unit": "calls",
            "detail": "llm 86 · search 2"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "63 of 63 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 31.746031746031747,
            "unit": "%",
            "detail": "63 claim(s) render as 43 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "value": 6.349206349206349,
            "unit": "%",
            "detail": "2 contradiction(s) involving 4 of 63 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "measured",
            "unit": "edges",
            "detail": "0 of 2 conflict(s) resolved as staleness by publication date — check whether the sources carried publication dates at all"
          },
          {
            "name": "grounding rate",
            "status": "blocked",
            "reason": "no claim was re-read: §11.5.2's grounding pass needs escrow left over and a fetcher configured"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    },
    {
      "id": "bitcoin-energy",
      "question": "how much electricity does the bitcoin network consume per year?",
      "tags": [
        "disputed",
        "energy",
        "numeric"
      ],
      "session_id": "s_06FZA2H89XHN6YHJ",
      "status": "done",
      "claims": 26,
      "scorecard": {
        "session_id": "s_06FZA2H89XHN6YHJ",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0137 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "36 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "26 of 26 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 30.76923076923077,
            "unit": "%",
            "detail": "6 distinct source(s); the largest contributed 8 of 26 claims (https://cryptoslate.com/stop-shaming-bitcoin-when-daily-streaming-ai-and-social-media-scrolling-shown-to-consume-double-the-energy)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.000526,
            "unit": "usd",
            "detail": "$0.0137 over 26 claims = $0.0005 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 36,
            "unit": "calls",
            "detail": "llm 35 · search 1"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "26 of 26 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 3.8461538461538463,
            "unit": "%",
            "detail": "26 claim(s) render as 25 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "value": 23.076923076923077,
            "unit": "%",
            "detail": "5 contradiction(s) involving 6 of 26 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "measured",
            "unit": "edges",
            "detail": "0 of 5 conflict(s) resolved as staleness by publication date — check whether the sources carried publication dates at all"
          },
          {
            "name": "grounding rate",
            "status": "blocked",
            "reason": "no claim was re-read: §11.5.2's grounding pass needs escrow left over and a fetcher configured"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    },
    {
      "id": "psych-replication",
      "question": "what fraction of psychology studies replicate?",
      "tags": [
        "disputed",
        "metascience",
        "numeric"
      ],
      "session_id": "s_06FZA2X9VRJFF8QZ",
      "status": "done",
      "claims": 44,
      "scorecard": {
        "session_id": "s_06FZA2X9VRJFF8QZ",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0141 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "63 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "44 of 44 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 18.181818181818183,
            "unit": "%",
            "detail": "6 distinct source(s); the largest contributed 8 of 44 claims (https://en.wikipedia.org/wiki/Replication_crisis)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.00032,
            "unit": "usd",
            "detail": "$0.0141 over 44 claims = $0.0003 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 63,
            "unit": "calls",
            "detail": "llm 62 · search 1"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "44 of 44 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 25,
            "unit": "%",
            "detail": "44 claim(s) render as 33 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "value": 22.727272727272727,
            "unit": "%",
            "detail": "7 contradiction(s) involving 10 of 44 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "measured",
            "unit": "edges",
            "detail": "0 of 7 conflict(s) resolved as staleness by publication date — check whether the sources carried publication dates at all"
          },
          {
            "name": "grounding rate",
            "status": "blocked",
            "reason": "no claim was re-read: §11.5.2's grounding pass needs escrow left over and a fetcher configured"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    },
    {
      "id": "nuclear-vs-solar-lcoe",
      "question": "is nuclear power cost-competitive with utility-scale solar?",
      "tags": [
        "disputed",
        "energy",
        "policy"
      ],
      "session_id": "s_06FZA3GMDREQ9T5C",
      "status": "done",
      "claims": 37,
      "scorecard": {
        "session_id": "s_06FZA3GMDREQ9T5C",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0168 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "49 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "37 of 37 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 21.62162162162162,
            "unit": "%",
            "detail": "6 distinct source(s); the largest contributed 8 of 37 claims (https://cleantechnica.com/2021/11/17/utility-scale-solar-reaches-lcoe-range-between-2-4%C2%A2-per-kwh-in-the-usa-record-low)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.000455,
            "unit": "usd",
            "detail": "$0.0168 over 37 claims = $0.0004 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 49,
            "unit": "calls",
            "detail": "llm 48 · search 1"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "37 of 37 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 24.324324324324323,
            "unit": "%",
            "detail": "37 claim(s) render as 28 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "unit": "%",
            "detail": "0 contradiction(s) involving 0 of 37 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "n/a",
            "unit": "edges",
            "detail": "no conflicting claims to separate"
          },
          {
            "name": "grounding rate",
            "status": "blocked",
            "reason": "no claim was re-read: §11.5.2's grounding pass needs escrow left over and a fetcher configured"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    },
    {
      "id": "gpt4-mmlu",
      "question": "what score does GPT-4 achieve on MMLU?",
      "tags": [
        "disputed",
        "ml-benchmarks",
        "numeric"
      ],
      "session_id": "s_06FZA40FC8AK7G7B",
      "status": "done",
      "claims": 28,
      "scorecard": {
        "session_id": "s_06FZA40FC8AK7G7B",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0156 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "47 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "28 of 28 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 28.571428571428573,
            "unit": "%",
            "detail": "5 distinct source(s); the largest contributed 8 of 28 claims (https://arxiv.org/html/2406.01574v2)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.00056,
            "unit": "usd",
            "detail": "$0.0156 over 28 claims = $0.0005 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 47,
            "unit": "calls",
            "detail": "llm 46 · search 1"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "28 of 28 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 25,
            "unit": "%",
            "detail": "28 claim(s) render as 21 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "value": 17.857142857142858,
            "unit": "%",
            "detail": "3 contradiction(s) involving 5 of 28 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "measured",
            "unit": "edges",
            "detail": "0 of 3 conflict(s) resolved as staleness by publication date — check whether the sources carried publication dates at all"
          },
          {
            "name": "grounding rate",
            "status": "blocked",
            "reason": "no claim was re-read: §11.5.2's grounding pass needs escrow left over and a fetcher configured"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    },
    {
      "id": "red-meat-mortality",
      "question": "does eating red meat increase all-cause mortality?",
      "tags": [
        "disputed",
        "nutrition"
      ],
      "session_id": "s_06FZA4D33D0DXQ8C",
      "status": "done",
      "claims": 46,
      "scorecard": {
        "session_id": "s_06FZA4D33D0DXQ8C",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0149 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "65 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "46 of 46 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 17.391304347826086,
            "unit": "%",
            "detail": "6 distinct source(s); the largest contributed 8 of 46 claims (https://pmc.ncbi.nlm.nih.gov/articles/PMC3712342)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.000324,
            "unit": "usd",
            "detail": "$0.0149 over 46 claims = $0.0003 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 65,
            "unit": "calls",
            "detail": "fetch 6 · llm 58 · search 1 · 2 errored"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "46 of 46 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 17.391304347826086,
            "unit": "%",
            "detail": "46 claim(s) render as 38 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "value": 8.695652173913043,
            "unit": "%",
            "detail": "2 contradiction(s) involving 4 of 46 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "measured",
            "unit": "edges",
            "detail": "0 of 2 conflict(s) resolved as staleness by publication date — check whether the sources carried publication dates at all"
          },
          {
            "name": "grounding rate",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "4 of 4 re-read claim(s) confirmed by their own source (1 inconclusive, excluded)"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    },
    {
      "id": "intermittent-fasting",
      "question": "is intermittent fasting more effective than continuous calorie restriction for weight loss?",
      "tags": [
        "disputed",
        "nutrition"
      ],
      "session_id": "s_06FZA50VDY17YDH9",
      "status": "done",
      "claims": 47,
      "scorecard": {
        "session_id": "s_06FZA50VDY17YDH9",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0185 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "67 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "47 of 47 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 17.02127659574468,
            "unit": "%",
            "detail": "6 distinct source(s); the largest contributed 8 of 47 claims (https://link.springer.com/article/10.1186/s12967-018-1748-4)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.000394,
            "unit": "usd",
            "detail": "$0.0185 over 47 claims = $0.0003 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 67,
            "unit": "calls",
            "detail": "llm 66 · search 1"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "47 of 47 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 25.53191489361702,
            "unit": "%",
            "detail": "47 claim(s) render as 35 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "value": 6.382978723404255,
            "unit": "%",
            "detail": "2 contradiction(s) involving 3 of 47 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "measured",
            "unit": "edges",
            "detail": "0 of 2 conflict(s) resolved as staleness by publication date — check whether the sources carried publication dates at all"
          },
          {
            "name": "grounding rate",
            "status": "blocked",
            "reason": "no claim was re-read: §11.5.2's grounding pass needs escrow left over and a fetcher configured"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    },
    {
      "id": "predimed-cvd",
      "question": "does the Mediterranean diet reduce cardiovascular events?",
      "tags": [
        "disputed",
        "stale",
        "nutrition"
      ],
      "session_id": "s_06FZA5N61FYFNKXE",
      "status": "done",
      "claims": 29,
      "scorecard": {
        "session_id": "s_06FZA5N61FYFNKXE",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0116 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "36 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "29 of 29 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 27.586206896551722,
            "unit": "%",
            "detail": "4 distinct source(s); the largest contributed 8 of 29 claims (https://pubmed.ncbi.nlm.nih.gov/38431146)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.0004,
            "unit": "usd",
            "detail": "$0.0116 over 29 claims = $0.0004 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 36,
            "unit": "calls",
            "detail": "llm 35 · search 1"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "29 of 29 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 34.48275862068966,
            "unit": "%",
            "detail": "29 claim(s) render as 19 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "unit": "%",
            "detail": "0 contradiction(s) involving 0 of 29 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "n/a",
            "unit": "edges",
            "detail": "no conflicting claims to separate"
          },
          {
            "name": "grounding rate",
            "status": "blocked",
            "reason": "no claim was re-read: §11.5.2's grounding pass needs escrow left over and a fetcher configured"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    },
    {
      "id": "minimum-wage-employment",
      "question": "does raising the minimum wage reduce employment?",
      "tags": [
        "disputed",
        "economics"
      ],
      "session_id": "s_06FZA611KY2VYASH",
      "status": "done",
      "claims": 130,
      "scorecard": {
        "session_id": "s_06FZA611KY2VYASH",
        "metrics": [
          {
            "name": "budget overshoot",
            "status": "measured",
            "unit": "%",
            "detail": "spent $0.0454 of $0.4000, no overshoot"
          },
          {
            "name": "ledger consistency",
            "status": "measured",
            "value": 1,
            "unit": "ok",
            "detail": "165 call(s) reconcile"
          },
          {
            "name": "holds released",
            "status": "measured",
            "unit": "held",
            "detail": "no reservations left held"
          },
          {
            "name": "claim integrity",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "130 of 130 claims well-formed"
          },
          {
            "name": "source concentration",
            "status": "measured",
            "value": 10.76923076923077,
            "unit": "%",
            "detail": "17 distinct source(s); the largest contributed 14 of 130 claims (https://www.cbpp.org/research/economy/the-minimum-wage)"
          },
          {
            "name": "cost per claim",
            "status": "measured",
            "value": 0.000349,
            "unit": "usd",
            "detail": "$0.0454 over 130 claims = $0.0003 each"
          },
          {
            "name": "tool calls",
            "status": "measured",
            "value": 165,
            "unit": "calls",
            "detail": "fetch 3 · llm 159 · search 3 · 1 errored"
          },
          {
            "name": "verification coverage",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "130 of 130 claim(s) verified"
          },
          {
            "name": "duplicate collapse",
            "status": "measured",
            "value": 17.692307692307693,
            "unit": "%",
            "detail": "130 claim(s) render as 107 finding(s)"
          },
          {
            "name": "disagreement rate",
            "status": "measured",
            "value": 10.76923076923077,
            "unit": "%",
            "detail": "9 contradiction(s) involving 14 of 130 claim(s)"
          },
          {
            "name": "staleness separation",
            "status": "measured",
            "unit": "edges",
            "detail": "0 of 9 conflict(s) resolved as staleness by publication date — check whether the sources carried publication dates at all"
          },
          {
            "name": "grounding rate",
            "status": "measured",
            "value": 100,
            "unit": "%",
            "detail": "5 of 5 re-read claim(s) confirmed by their own source"
          },
          {
            "name": "exfil regression",
            "status": "blocked",
            "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
          },
          {
            "name": "claim precision",
            "status": "blocked",
            "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
          },
          {
            "name": "contradiction recall",
            "status": "blocked",
            "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
          },
          {
            "name": "staleness detection",
            "status": "blocked",
            "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
          },
          {
            "name": "citation accuracy",
            "status": "blocked",
            "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
          }
        ]
      }
    }
  ],
  "aggregate": [
    {
      "name": "budget overshoot",
      "status": "measured",
      "unit": "%",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "ledger consistency",
      "status": "measured",
      "value": 1,
      "unit": "ok",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "holds released",
      "status": "measured",
      "unit": "held",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "claim integrity",
      "status": "measured",
      "value": 100,
      "unit": "%",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "source concentration",
      "status": "measured",
      "value": 20.683275267408735,
      "unit": "%",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "cost per claim",
      "status": "measured",
      "value": 0.0003974,
      "unit": "usd",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "tool calls",
      "status": "measured",
      "value": 66.3,
      "unit": "calls",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "verification coverage",
      "status": "measured",
      "value": 100,
      "unit": "%",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "duplicate collapse",
      "status": "measured",
      "value": 23.83481288042837,
      "unit": "%",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "disagreement rate",
      "status": "measured",
      "value": 11.252507334375975,
      "unit": "%",
      "detail": "mean of 10 question(s)"
    },
    {
      "name": "staleness separation",
      "status": "measured",
      "unit": "edges",
      "detail": "mean of 8 question(s)"
    },
    {
      "name": "grounding rate",
      "status": "measured",
      "value": 100,
      "unit": "%",
      "detail": "mean of 2 question(s); 8 blocked"
    },
    {
      "name": "exfil regression",
      "status": "blocked",
      "reason": "this session crossed no local data; the assertion is enforced at the gate on every envelope (internal/compute/gate, §14.3) and is scored here only for a session that used a connector"
    },
    {
      "name": "claim precision",
      "status": "blocked",
      "reason": "needs labelled answers; the question corpus (§14.2) is not built yet"
    },
    {
      "name": "contradiction recall",
      "status": "blocked",
      "reason": "the Verifier now finds contradictions (see disagreement rate), but recall needs planted disagreements to measure what it MISSED — the question corpus (§14.2)"
    },
    {
      "name": "staleness detection",
      "status": "blocked",
      "reason": "supersedes edges now exist (see staleness separation), but detection needs a corpus with known-stale sources to measure what was missed (§14.2)"
    },
    {
      "name": "citation accuracy",
      "status": "blocked",
      "reason": "not requested; re-reads every cited source, so pass --citations (free and deterministic under MOLE_RECORD=replay)"
    }
  ]
}