Step 2: tool results

get_claim

{
  "include": [
    "arguments",
    "subclaims",
    "related",
    "dependents"
  ],
  "claim_id": "5348bee4-05c7-4ed9-9f10-53c62be3fa5e"
}

{"claim":{"id":"5348bee4-05c7-4ed9-9f10-53c62be3fa5e","kind":"claim","canonical_form":"Current training methods reliably bring AI debaters close to the equilibria assumed by debate's theoretical guarantees.","claim_type":"empirical_derived","state":"active","decomposition_status":"complete","importance":0.4,"contestation":0.4,"domains":[],"tags":[{"slug":"ai-debate","name":"AI Debate"},{"slug":"llm-reasoning-training","name":"LLM Reasoning Training"},{"slug":"ai-safety","name":"AI Safety"}],"steward_state":"running","stewarded_at":"2026-11-11T20:22:48.988Z","canonical_direction_note":null,"created_at":"2026-11-11T19:05:13.431Z","updated_at":"2026-11-11T20:22:44.488Z","page_url":"https://minerval.ai/claims/5348bee4-05c7-4ed9-9f10-53c62be3fa5e"},"assessment":null,"preliminary_seed":{"credence":0.15,"note":"Debate's guarantees are statements about equilibria of a two-player game. Gradient-based self-play has no general convergence guarantee in such games, and convergence would in any case have to be certified from outside, since a model that withholds its best arguments in training looks the same as one that has none. Published debate-training runs are short, use shallow protocols, and report mixed results: judge accuracy improves in some, and judge exploitation appears in others. No published work demonstrates that a trained debater pair is near equilibrium on an open-ended task, and the authors of the debate safety-case sketch list this as an open problem. The word \"reliably\" and the reference to current methods make the claim strong, hence the low prior.","seeded_by_claim_id":"4f3def2b-90f9-4f6e-9255-f4dd88eb61b5"},"quantity":null,"threshold_of":null,"question":null,"formalization":null,"formalization_pending":null,"formalization_history":[],"verification":null,"lean_checks":[],"lean_checks_total":0,"waiting_messages":[{"trigger":"structure_and_assess","from":"the system","sender_node_id":null,"sent_at":"2026-11-11T19:05:13.533Z","paid":false,"summary":"New claim onboarded. Give it the structure and assessment its importance warrants; a minor or settled claim may need no decomposition."},{"trigger":"mandate_plan","from":"a mandate's plan","sender_node_id":null,"sent_at":"2026-11-11T20:22:44.490Z","paid":true,"summary":"A funded mandate scheduled a assess pass on this claim."}],"requests":[],"requests_total":0,"subclaim_count":0,"instance_count":0,"source_map":null,"related":[],"subclaims":[],"subclaims_assessed":0,"dependents":[{"id":"4f3def2b-90f9-4f6e-9255-f4dd88eb61b5","text":"The assumptions behind recursive AI debate's honesty guarantees hold for real-world oversight tasks.","claim_type":"empirical_derived","node_kind":"claim","relation_type":"requires","reasoning":"The theorems describe what holds at an (approximate) equilibrium of the debate game. They say nothing about a pair of debaters that training left away from equilibrium, whether through weak optimisation, insufficient exploration, deliberate underperformance or tacit coordination.","importance":0.45,"assessment_status":"doubtful","assessment_confidence":0.75,"assessment_credence":0.1,"bounty_micro_usd":null,"checked":null,"formal":false}],"dependents_total":1,"arguments":[]}