{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:JYYNUACG4T7G7T43ZVFUACSCIQ","short_pith_number":"pith:JYYNUACG","canonical_record":{"source":{"id":"2409.16636","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-09-25T05:28:33Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"56c1706a48997e0a32608142438dbc436c6d0c17d8b652fd0c4961b01f70cce4","abstract_canon_sha256":"39666c73165f00d6f119e22850e34d2a6eabd8ddd50c2e93b086738aae77ff06"},"schema_version":"1.0"},"canonical_sha256":"4e30da0046e4fe6fcf9bcd4b400a4244382691a4c68b4915caa641ce51d4dd31","source":{"kind":"arxiv","id":"2409.16636","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2409.16636","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"arxiv_version","alias_value":"2409.16636v1","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.16636","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"pith_short_12","alias_value":"JYYNUACG4T7G","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"pith_short_16","alias_value":"JYYNUACG4T7G7T43","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"pith_short_8","alias_value":"JYYNUACG","created_at":"2026-07-05T09:11:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:JYYNUACG4T7G7T43ZVFUACSCIQ","target":"record","payload":{"canonical_record":{"source":{"id":"2409.16636","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-09-25T05:28:33Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"56c1706a48997e0a32608142438dbc436c6d0c17d8b652fd0c4961b01f70cce4","abstract_canon_sha256":"39666c73165f00d6f119e22850e34d2a6eabd8ddd50c2e93b086738aae77ff06"},"schema_version":"1.0"},"canonical_sha256":"4e30da0046e4fe6fcf9bcd4b400a4244382691a4c68b4915caa641ce51d4dd31","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:11:37.100462Z","signature_b64":"HjF3pbw8/E0ER0b5pe2xBkxeP3kHiKwp4RsB93IDA/uhntx9BgWc3G6mEfupRQpG+otUodzhtV7s1CDS2nfRBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4e30da0046e4fe6fcf9bcd4b400a4244382691a4c68b4915caa641ce51d4dd31","last_reissued_at":"2026-07-05T09:11:37.099938Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:11:37.099938Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2409.16636","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T09:11:37Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/tFrYbYdseG9L+VeRmV8CnRNREfp4xn0SBYZyCERPngePRObpL88WnDPkEj+XMr9TJG9ZqWStSgECsNyjcOjAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T07:21:43.049712Z"},"content_sha256":"f2bd55b212a91930e3f0c946d0691ab558cf91f7333f4525b09d4ba8d71fd98f","schema_version":"1.0","event_id":"sha256:f2bd55b212a91930e3f0c946d0691ab558cf91f7333f4525b09d4ba8d71fd98f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:JYYNUACG4T7G7T43ZVFUACSCIQ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Training Language Models to Win Debates with Self-Play Improves Judge Accuracy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"David Rein, Julian Michael, Samuel Arnesen","submitted_at":"2024-09-25T05:28:33Z","abstract_excerpt":"We test the robustness of debate as a method of scalable oversight by training models to debate with data generated via self-play. In a long-context reading comprehension task, we find that language model based evaluators answer questions more accurately when judging models optimized to win debates. By contrast, we find no such relationship for consultancy models trained to persuade a judge without an opposing debater present. In quantitative and qualitative comparisons between our debate models and novel consultancy baselines, we find evidence that debate training encourages stronger and more"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.16636","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2409.16636/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T09:11:37Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"jlIdlSiAzjyX3Fz80aTxUup4vTfpkqafLctR2/k/B37W3oC0GjkEaBc1io7dgLR+MbaehY4bM/AbCbLMaDznDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T07:21:43.050516Z"},"content_sha256":"b046c93cecfcf88774a7d29c57b9ca407b379819251666b37a40afa06ae28c6d","schema_version":"1.0","event_id":"sha256:b046c93cecfcf88774a7d29c57b9ca407b379819251666b37a40afa06ae28c6d"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/JYYNUACG4T7G7T43ZVFUACSCIQ/bundle.json","state_url":"https://pith.science/pith/JYYNUACG4T7G7T43ZVFUACSCIQ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/JYYNUACG4T7G7T43ZVFUACSCIQ/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-18T07:21:43Z","links":{"resolver":"https://pith.science/pith/JYYNUACG4T7G7T43ZVFUACSCIQ","bundle":"https://pith.science/pith/JYYNUACG4T7G7T43ZVFUACSCIQ/bundle.json","state":"https://pith.science/pith/JYYNUACG4T7G7T43ZVFUACSCIQ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/JYYNUACG4T7G7T43ZVFUACSCIQ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:JYYNUACG4T7G7T43ZVFUACSCIQ","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"39666c73165f00d6f119e22850e34d2a6eabd8ddd50c2e93b086738aae77ff06","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-09-25T05:28:33Z","title_canon_sha256":"56c1706a48997e0a32608142438dbc436c6d0c17d8b652fd0c4961b01f70cce4"},"schema_version":"1.0","source":{"id":"2409.16636","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2409.16636","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"arxiv_version","alias_value":"2409.16636v1","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.16636","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"pith_short_12","alias_value":"JYYNUACG4T7G","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"pith_short_16","alias_value":"JYYNUACG4T7G7T43","created_at":"2026-07-05T09:11:37Z"},{"alias_kind":"pith_short_8","alias_value":"JYYNUACG","created_at":"2026-07-05T09:11:37Z"}],"graph_snapshots":[{"event_id":"sha256:b046c93cecfcf88774a7d29c57b9ca407b379819251666b37a40afa06ae28c6d","target":"graph","created_at":"2026-07-05T09:11:37Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2409.16636/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We test the robustness of debate as a method of scalable oversight by training models to debate with data generated via self-play. In a long-context reading comprehension task, we find that language model based evaluators answer questions more accurately when judging models optimized to win debates. By contrast, we find no such relationship for consultancy models trained to persuade a judge without an opposing debater present. In quantitative and qualitative comparisons between our debate models and novel consultancy baselines, we find evidence that debate training encourages stronger and more","authors_text":"David Rein, Julian Michael, Samuel Arnesen","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-09-25T05:28:33Z","title":"Training Language Models to Win Debates with Self-Play Improves Judge Accuracy"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.16636","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f2bd55b212a91930e3f0c946d0691ab558cf91f7333f4525b09d4ba8d71fd98f","target":"record","created_at":"2026-07-05T09:11:37Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"39666c73165f00d6f119e22850e34d2a6eabd8ddd50c2e93b086738aae77ff06","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-09-25T05:28:33Z","title_canon_sha256":"56c1706a48997e0a32608142438dbc436c6d0c17d8b652fd0c4961b01f70cce4"},"schema_version":"1.0","source":{"id":"2409.16636","kind":"arxiv","version":1}},"canonical_sha256":"4e30da0046e4fe6fcf9bcd4b400a4244382691a4c68b4915caa641ce51d4dd31","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4e30da0046e4fe6fcf9bcd4b400a4244382691a4c68b4915caa641ce51d4dd31","first_computed_at":"2026-07-05T09:11:37.099938Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:11:37.099938Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"HjF3pbw8/E0ER0b5pe2xBkxeP3kHiKwp4RsB93IDA/uhntx9BgWc3G6mEfupRQpG+otUodzhtV7s1CDS2nfRBg==","signature_status":"signed_v1","signed_at":"2026-07-05T09:11:37.100462Z","signed_message":"canonical_sha256_bytes"},"source_id":"2409.16636","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f2bd55b212a91930e3f0c946d0691ab558cf91f7333f4525b09d4ba8d71fd98f","sha256:b046c93cecfcf88774a7d29c57b9ca407b379819251666b37a40afa06ae28c6d"],"state_sha256":"eb97cdd46960f60fd977f2b24fcfac82ff62762f648238e4c76a04cee488c286"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"fSBYY5ueBKQGEd8ntGIQC6KaYp89j3xWgSZUJ2ribBG+D4e/6NqKdgwBOYnj3GaRUeFx47aeVO6ISnwq+TKhCw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-18T07:21:43.065398Z","bundle_sha256":"49076a84f21c6a1bcecc5c49e9f349c9c40e98a48f4f1cf8aed027998a0764b0"}}