{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:CEGTEPVJ35YBGDPQGU73H6QQJJ","short_pith_number":"pith:CEGTEPVJ","canonical_record":{"source":{"id":"2508.21787","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-29T17:03:47Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d4eb40fad596d1a0a907358e2c298c3e43fadd343b6929fd595d78c586929333","abstract_canon_sha256":"e44f8eea31d98b461605e3f2ff11d0389556a1faf74a1a2d317edf259c145676"},"schema_version":"1.0"},"canonical_sha256":"110d323ea9df70130df0353fb3fa104a6c052c984d7456a16fc1b4eee68d4bc8","source":{"kind":"arxiv","id":"2508.21787","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.21787","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"arxiv_version","alias_value":"2508.21787v3","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.21787","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"pith_short_12","alias_value":"CEGTEPVJ35YB","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"pith_short_16","alias_value":"CEGTEPVJ35YBGDPQ","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"pith_short_8","alias_value":"CEGTEPVJ","created_at":"2026-07-14T02:22:28Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:CEGTEPVJ35YBGDPQGU73H6QQJJ","target":"record","payload":{"canonical_record":{"source":{"id":"2508.21787","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-29T17:03:47Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d4eb40fad596d1a0a907358e2c298c3e43fadd343b6929fd595d78c586929333","abstract_canon_sha256":"e44f8eea31d98b461605e3f2ff11d0389556a1faf74a1a2d317edf259c145676"},"schema_version":"1.0"},"canonical_sha256":"110d323ea9df70130df0353fb3fa104a6c052c984d7456a16fc1b4eee68d4bc8","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-14T02:22:28.667711Z","signature_b64":"n1BFqzL5wVocnOejmJyuXKuhqwutFG59DM7QxxDIS01VG5Q30imK9X3cOcT3pIxo1oK/U8EmujKhobzuN63dAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"110d323ea9df70130df0353fb3fa104a6c052c984d7456a16fc1b4eee68d4bc8","last_reissued_at":"2026-07-14T02:22:28.666784Z","signature_status":"signed_v1","first_computed_at":"2026-07-14T02:22:28.666784Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2508.21787","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-14T02:22:28Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8cA6e3FxxBB0cL91z7lxN4LxQ0wAnQSV/JckUM0qApLgVTs9YQ3ARMAx9dcjGeesCuJ3REtvacZM/Bo2fuDCBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T04:54:28.782924Z"},"content_sha256":"1e00a29772b506a6383e2da8844072bf961497ad3da9e436b2ca82f9d22b0cd0","schema_version":"1.0","event_id":"sha256:1e00a29772b506a6383e2da8844072bf961497ad3da9e436b2ca82f9d22b0cd0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:CEGTEPVJ35YBGDPQGU73H6QQJJ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"PiCSAR: Probabilistic Confidence Selection And Ranking for Reasoning Chains","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A training-free score based on the joint log-likelihood of reasoning steps and final answer selects correct chains more reliably than prior methods.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Aryo Pradipta Gema, Eleonora Giunchiglia, Joshua Ong Jun Leang, Pasquale Minervini, Shay B. Cohen, Sohee Yang, Wai-Chung Kwan, Wenda Li, Xuanli He, Zheng Zhao","submitted_at":"2025-08-29T17:03:47Z","abstract_excerpt":"Best-of-n sampling improves the accuracy of large language models (LLMs) and large reasoning models (LRMs) by generating multiple candidate solutions and selecting the one with the highest reward. The key challenge for reasoning tasks is designing a scoring function that can identify correct reasoning chains without access to ground-truth answers. We propose Probabilistic Confidence Selection And Ranking (PiCSAR): a simple, training-free method that scores each candidate generation using the joint log-likelihood of the reasoning and final answer. The joint log-likelihood of the reasoning and f"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"PiCSAR achieves substantial gains across diverse benchmarks (+10.18 on MATH500, +9.81 on AIME2025), outperforming baselines with at least 2x fewer samples in 16 out of 20 comparisons.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"Correct reasoning chains exhibit significantly higher reasoning and answer confidence (as measured by joint log-likelihood) than incorrect ones, allowing the scoring function to identify them reliably.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"PiCSAR ranks reasoning chains by joint log-likelihood of reasoning and answer to select better outputs, delivering large gains on math benchmarks with reduced sampling.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A training-free score based on the joint log-likelihood of reasoning steps and final answer selects correct chains more reliably than prior methods.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"b7b75f23240e20ed37108eb6b05d6b49af430039bd767536bdc9ec90b9ffe450"},"source":{"id":"2508.21787","kind":"arxiv","version":3},"verdict":{"id":"905bf173-265e-4083-a4e5-147848871dff","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-18T20:02:00.204512Z","strongest_claim":"PiCSAR achieves substantial gains across diverse benchmarks (+10.18 on MATH500, +9.81 on AIME2025), outperforming baselines with at least 2x fewer samples in 16 out of 20 comparisons.","one_line_summary":"PiCSAR ranks reasoning chains by joint log-likelihood of reasoning and answer to select better outputs, delivering large gains on math benchmarks with reduced sampling.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"Correct reasoning chains exhibit significantly higher reasoning and answer confidence (as measured by joint log-likelihood) than incorrect ones, allowing the scoring function to identify them reliably.","pith_extraction_headline":"A training-free score based on the joint log-likelihood of reasoning steps and final answer selects correct chains more reliably than prior methods."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.21787/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":2,"snapshot_sha256":"098a790fb1a53bea0e4412e74b9cddfabc14bc57cd79c779e52603a2e8c17f9f"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"905bf173-265e-4083-a4e5-147848871dff"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-14T02:22:28Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"bEYm2iV7DW9x7opkeD1iLSAKsNQyZOMkGuy5ZR8dkdNYFeFZTjSrYz/86ZoLSjXcGbJYCL9BNSpS2Nt4GbOMAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T04:54:28.783691Z"},"content_sha256":"007c5bcdb12c65760da048f286fdef0e87576ee96c8b7535315142786465630e","schema_version":"1.0","event_id":"sha256:007c5bcdb12c65760da048f286fdef0e87576ee96c8b7535315142786465630e"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/CEGTEPVJ35YBGDPQGU73H6QQJJ/bundle.json","state_url":"https://pith.science/pith/CEGTEPVJ35YBGDPQGU73H6QQJJ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/CEGTEPVJ35YBGDPQGU73H6QQJJ/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":"2026-08-04T04:54:28Z","links":{"resolver":"https://pith.science/pith/CEGTEPVJ35YBGDPQGU73H6QQJJ","bundle":"https://pith.science/pith/CEGTEPVJ35YBGDPQGU73H6QQJJ/bundle.json","state":"https://pith.science/pith/CEGTEPVJ35YBGDPQGU73H6QQJJ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/CEGTEPVJ35YBGDPQGU73H6QQJJ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:CEGTEPVJ35YBGDPQGU73H6QQJJ","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":"e44f8eea31d98b461605e3f2ff11d0389556a1faf74a1a2d317edf259c145676","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-29T17:03:47Z","title_canon_sha256":"d4eb40fad596d1a0a907358e2c298c3e43fadd343b6929fd595d78c586929333"},"schema_version":"1.0","source":{"id":"2508.21787","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.21787","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"arxiv_version","alias_value":"2508.21787v3","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.21787","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"pith_short_12","alias_value":"CEGTEPVJ35YB","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"pith_short_16","alias_value":"CEGTEPVJ35YBGDPQ","created_at":"2026-07-14T02:22:28Z"},{"alias_kind":"pith_short_8","alias_value":"CEGTEPVJ","created_at":"2026-07-14T02:22:28Z"}],"graph_snapshots":[{"event_id":"sha256:007c5bcdb12c65760da048f286fdef0e87576ee96c8b7535315142786465630e","target":"graph","created_at":"2026-07-14T02:22:28Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"PiCSAR achieves substantial gains across diverse benchmarks (+10.18 on MATH500, +9.81 on AIME2025), outperforming baselines with at least 2x fewer samples in 16 out of 20 comparisons."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"Correct reasoning chains exhibit significantly higher reasoning and answer confidence (as measured by joint log-likelihood) than incorrect ones, allowing the scoring function to identify them reliably."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"PiCSAR ranks reasoning chains by joint log-likelihood of reasoning and answer to select better outputs, delivering large gains on math benchmarks with reduced sampling."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A training-free score based on the joint log-likelihood of reasoning steps and final answer selects correct chains more reliably than prior methods."}],"snapshot_sha256":"b7b75f23240e20ed37108eb6b05d6b49af430039bd767536bdc9ec90b9ffe450"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"098a790fb1a53bea0e4412e74b9cddfabc14bc57cd79c779e52603a2e8c17f9f"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2508.21787/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Best-of-n sampling improves the accuracy of large language models (LLMs) and large reasoning models (LRMs) by generating multiple candidate solutions and selecting the one with the highest reward. The key challenge for reasoning tasks is designing a scoring function that can identify correct reasoning chains without access to ground-truth answers. We propose Probabilistic Confidence Selection And Ranking (PiCSAR): a simple, training-free method that scores each candidate generation using the joint log-likelihood of the reasoning and final answer. The joint log-likelihood of the reasoning and f","authors_text":"Aryo Pradipta Gema, Eleonora Giunchiglia, Joshua Ong Jun Leang, Pasquale Minervini, Shay B. Cohen, Sohee Yang, Wai-Chung Kwan, Wenda Li, Xuanli He, Zheng Zhao","cross_cats":["cs.AI"],"headline":"A training-free score based on the joint log-likelihood of reasoning steps and final answer selects correct chains more reliably than prior methods.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-29T17:03:47Z","title":"PiCSAR: Probabilistic Confidence Selection And Ranking for Reasoning Chains"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.21787","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-18T20:02:00.204512Z","id":"905bf173-265e-4083-a4e5-147848871dff","model_set":{"reader":"grok-4.3"},"one_line_summary":"PiCSAR ranks reasoning chains by joint log-likelihood of reasoning and answer to select better outputs, delivering large gains on math benchmarks with reduced sampling.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A training-free score based on the joint log-likelihood of reasoning steps and final answer selects correct chains more reliably than prior methods.","strongest_claim":"PiCSAR achieves substantial gains across diverse benchmarks (+10.18 on MATH500, +9.81 on AIME2025), outperforming baselines with at least 2x fewer samples in 16 out of 20 comparisons.","weakest_assumption":"Correct reasoning chains exhibit significantly higher reasoning and answer confidence (as measured by joint log-likelihood) than incorrect ones, allowing the scoring function to identify them reliably."}},"verdict_id":"905bf173-265e-4083-a4e5-147848871dff"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:1e00a29772b506a6383e2da8844072bf961497ad3da9e436b2ca82f9d22b0cd0","target":"record","created_at":"2026-07-14T02:22:28Z","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":"e44f8eea31d98b461605e3f2ff11d0389556a1faf74a1a2d317edf259c145676","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-29T17:03:47Z","title_canon_sha256":"d4eb40fad596d1a0a907358e2c298c3e43fadd343b6929fd595d78c586929333"},"schema_version":"1.0","source":{"id":"2508.21787","kind":"arxiv","version":3}},"canonical_sha256":"110d323ea9df70130df0353fb3fa104a6c052c984d7456a16fc1b4eee68d4bc8","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"110d323ea9df70130df0353fb3fa104a6c052c984d7456a16fc1b4eee68d4bc8","first_computed_at":"2026-07-14T02:22:28.666784Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-14T02:22:28.666784Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"n1BFqzL5wVocnOejmJyuXKuhqwutFG59DM7QxxDIS01VG5Q30imK9X3cOcT3pIxo1oK/U8EmujKhobzuN63dAQ==","signature_status":"signed_v1","signed_at":"2026-07-14T02:22:28.667711Z","signed_message":"canonical_sha256_bytes"},"source_id":"2508.21787","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:1e00a29772b506a6383e2da8844072bf961497ad3da9e436b2ca82f9d22b0cd0","sha256:007c5bcdb12c65760da048f286fdef0e87576ee96c8b7535315142786465630e"],"state_sha256":"b89e9236b29ab6a2c395836abfd4a7882868f4b6537e27cfbc8be3ab5b7390b3"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"J62OrCGT1pnCcpIRTyzkn7WWHvVsfFTKPdK9poT/Y/Sa8TJBdSkqnWN01P/s2SXMTPzGo5F1XrVP8hPlDIg9Dw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T04:54:28.789387Z","bundle_sha256":"2a9122a98f8fa6119c905dcfd768f2644c0274fe008bed92211e0458d3ccf324"}}