{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:F3GCK6BL4DCIKQKY65LB4W6SMP","short_pith_number":"pith:F3GCK6BL","schema_version":"1.0","canonical_sha256":"2ecc25782be0c4854158f7561e5bd263e6dd257fd53bcf47669d9e9f17f877f6","source":{"kind":"arxiv","id":"2501.13007","version":2},"attestation_state":"computed","paper":{"title":"PairJudge RM: Perform Best-of-N Sampling with Knockout Tournament","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Juanzi Li, Lei Hou, Rui Min, Yantao Liu, Yixin Cao, Zijun Yao","submitted_at":"2025-01-22T16:49:37Z","abstract_excerpt":"Best-of-N (BoN) sampling, a common strategy for test-time scaling of Large Language Models (LLMs), relies on reward models to select the best candidate solution from multiple generations. However, traditional reward models often assign arbitrary and inconsistent scores, limiting their effectiveness. To address this, we propose a Pairwise Judge Reward Model (PariJudge RM) combined with a knockout tournament for BoN sampling. Instead of assigning absolute scores, given one math problem, PariJudge RM judges two candidate solutions' correctness with chain-of-thought reasoning simultaneously. This "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2501.13007","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-22T16:49:37Z","cross_cats_sorted":[],"title_canon_sha256":"9d92268e5dfe1d60bce843745a595ef0baa6873ef86f0cda29528cc5b46e3625","abstract_canon_sha256":"59ff0f3aef01faca64a7cbe9e82a96286792025bd1de92d5e0280c5564afce5f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:16:41.075603Z","signature_b64":"eTgRNaEDd5WiHPXk/9zQXl2Xx400jvrv8HJ0i2hOOLUNYoPyTMhq4gikrshTk2OtoaifUOM0Sl3nLXiSsVtwBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2ecc25782be0c4854158f7561e5bd263e6dd257fd53bcf47669d9e9f17f877f6","last_reissued_at":"2026-07-05T10:16:41.075174Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:16:41.075174Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PairJudge RM: Perform Best-of-N Sampling with Knockout Tournament","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Juanzi Li, Lei Hou, Rui Min, Yantao Liu, Yixin Cao, Zijun Yao","submitted_at":"2025-01-22T16:49:37Z","abstract_excerpt":"Best-of-N (BoN) sampling, a common strategy for test-time scaling of Large Language Models (LLMs), relies on reward models to select the best candidate solution from multiple generations. However, traditional reward models often assign arbitrary and inconsistent scores, limiting their effectiveness. To address this, we propose a Pairwise Judge Reward Model (PariJudge RM) combined with a knockout tournament for BoN sampling. Instead of assigning absolute scores, given one math problem, PariJudge RM judges two candidate solutions' correctness with chain-of-thought reasoning simultaneously. This "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.13007","kind":"arxiv","version":2},"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/2501.13007/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2501.13007","created_at":"2026-07-05T10:16:41.075242+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.13007v2","created_at":"2026-07-05T10:16:41.075242+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.13007","created_at":"2026-07-05T10:16:41.075242+00:00"},{"alias_kind":"pith_short_12","alias_value":"F3GCK6BL4DCI","created_at":"2026-07-05T10:16:41.075242+00:00"},{"alias_kind":"pith_short_16","alias_value":"F3GCK6BL4DCIKQKY","created_at":"2026-07-05T10:16:41.075242+00:00"},{"alias_kind":"pith_short_8","alias_value":"F3GCK6BL","created_at":"2026-07-05T10:16:41.075242+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09076","citing_title":"Beyond Scalar Rewards by Internalizing Reasoning into Score Distributions","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2510.14232","citing_title":"Scaling Test-Time Compute to Achieve IOI Gold Medal with Open-Weight Models","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02507","citing_title":"Reinforcement Learning from Human Feedback: A Statistical Perspective","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12384","citing_title":"Scalable Token-Level Hallucination Detection in Large Language Models","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06219","citing_title":"Joint Consistency: A Unified Test-Time Aggregation Framework via Energy Minimization","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/F3GCK6BL4DCIKQKY65LB4W6SMP","json":"https://pith.science/pith/F3GCK6BL4DCIKQKY65LB4W6SMP.json","graph_json":"https://pith.science/api/pith-number/F3GCK6BL4DCIKQKY65LB4W6SMP/graph.json","events_json":"https://pith.science/api/pith-number/F3GCK6BL4DCIKQKY65LB4W6SMP/events.json","paper":"https://pith.science/paper/F3GCK6BL"},"agent_actions":{"view_html":"https://pith.science/pith/F3GCK6BL4DCIKQKY65LB4W6SMP","download_json":"https://pith.science/pith/F3GCK6BL4DCIKQKY65LB4W6SMP.json","view_paper":"https://pith.science/paper/F3GCK6BL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.13007&json=true","fetch_graph":"https://pith.science/api/pith-number/F3GCK6BL4DCIKQKY65LB4W6SMP/graph.json","fetch_events":"https://pith.science/api/pith-number/F3GCK6BL4DCIKQKY65LB4W6SMP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F3GCK6BL4DCIKQKY65LB4W6SMP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F3GCK6BL4DCIKQKY65LB4W6SMP/action/storage_attestation","attest_author":"https://pith.science/pith/F3GCK6BL4DCIKQKY65LB4W6SMP/action/author_attestation","sign_citation":"https://pith.science/pith/F3GCK6BL4DCIKQKY65LB4W6SMP/action/citation_signature","submit_replication":"https://pith.science/pith/F3GCK6BL4DCIKQKY65LB4W6SMP/action/replication_record"}},"created_at":"2026-07-05T10:16:41.075242+00:00","updated_at":"2026-07-05T10:16:41.075242+00:00"}