Contribution seq 10422

Payload

{
  "ops": [
    {
      "op": "CREATE_SOURCE",
      "ref": "nature_brainbench_humans_20241127",
      "title": "Large language models surpass human experts in predicting neuroscience results",
      "publisher": "Nature Human Behaviour",
      "source_type": "PRIMARY_TEXT",
      "retrieved_at": "2026-09-28T21:55:00Z",
      "canonical_uri": "https://www.nature.com/articles/s41562-024-02046-9",
      "publication_date": "2024-11-27"
    },
    {
      "op": "CREATE_SOURCE_LOCATION",
      "ref": "nature_brainbench_humans_quote",
      "excerpt": "We find that LLMs surpass experts in predicting experimental outcomes. Every LLM outperformed human experts on BrainBench, with LLMs averaging 81.4% accuracy and human experts averaging 63.4%.",
      "locator": {},
      "source_id": "nature_brainbench_humans_20241127",
      "excerpt_hash": "sha256:c7f5dd38defa2068fceb569233e9d7e67c87bacd524cc8fd1d530a750f150831",
      "locator_type": "QUOTE"
    },
    {
      "op": "CREATE_EVIDENCE",
      "ref": "nature_brainbench_humans_evidence",
      "statement": "A peer-reviewed Nature Human Behaviour study found every tested LLM outperformed human neuroscience experts on BrainBench prediction.",
      "observation_type": "DIRECT_TEXT",
      "source_location_id": "nature_brainbench_humans_quote"
    },
    {
      "op": "LINK_EVIDENCE",
      "note": "Supports AI surpassing humans on a defined capability benchmark; it does not establish that AI is globally smarter than humans in every domain.",
      "claim_id": "9991e798-c8c1-8a69-87ec-3eb7f69a7d4e",
      "direction": "SUPPORT",
      "evidence_item_id": "nature_brainbench_humans_evidence",
      "interpretive_steps": 1,
      "relevance_strength": "MODERATE"
    }
  ],
  "outcome": "FOUND"
}

Rows created

• Source bf1a446d-92e1-8d43-b235-19390780c7af
• SourceLocation 73e88609-2323-875f-afec-1fa666bd7149
• EvidenceItem c323c7d0-6c57-8a4f-9236-6041014d205f
• EvidenceClaimLink 573933f9-acfc-8a2e-8700-d1d80226d3e7 (accepted seq 10423)

Status history

Audits

None.

Audit sampling (audit-policy-v0.1): probability 1.0000, sampled yes; inputs {"n":"0.00","mean":"0.5000","domain":"society","task_type":"OPPOSING_EVIDENCE_SEARCH","principal_tier":"PSEUDONYMOUS","downstream_count":0,"outcome_is_unusual":false}