{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:X2ICAZ3NMGW5BUHCYHKPFODSRA","short_pith_number":"pith:X2ICAZ3N","schema_version":"1.0","canonical_sha256":"be9020676d61add0d0e2c1d4f2b872882bb27660b9e8e805948f138eab6a5cb2","source":{"kind":"arxiv","id":"2602.03061","version":2},"attestation_state":"computed","paper":{"title":"Evaluating LLMs When They Do Not Know the Answer: Statistical Evaluation of Mathematical Reasoning via Comparative Signals","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","math.ST","stat.ME","stat.ML","stat.TH"],"primary_cat":"cs.LG","authors_text":"Can Jin, Linjun Zhang, Ruijia Wu, Yang Zhou, Zhixian Zhang, Zihan Dong","submitted_at":"2026-02-03T03:40:01Z","abstract_excerpt":"Evaluating mathematical reasoning in LLMs is constrained by limited benchmark sizes and inherent model stochasticity, yielding high-variance accuracy estimates and unstable rankings across platforms. On difficult problems, an LLM may fail to produce a correct final answer, yet still provide reliable pairwise comparison signals indicating which of two candidate solutions is better. We leverage this observation to design a statistically efficient evaluation framework that combines standard labeled outcomes with pairwise comparison signals obtained by having models judge auxiliary reasoning chain"},"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":"2602.03061","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-02-03T03:40:01Z","cross_cats_sorted":["cs.AI","math.ST","stat.ME","stat.ML","stat.TH"],"title_canon_sha256":"aa640b3a4ebe50756c5935442b6cd6b1743d5e8f701d807bd538d7d1c96aaf47","abstract_canon_sha256":"83c532e906bb05d113776586f393d0e105071ef4b17c3a77483c94b1223a2ba4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-21T01:21:42.828114Z","signature_b64":"ID1p6IlZq8IKkQvphCSGsQD717f1Rvg4d/Bu+FbYiE/HgPH1n/UL36IdINlsHAE8b87QKWVxi7bnXh2aTsmhCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"be9020676d61add0d0e2c1d4f2b872882bb27660b9e8e805948f138eab6a5cb2","last_reissued_at":"2026-07-21T01:21:42.827160Z","signature_status":"signed_v1","first_computed_at":"2026-07-21T01:21:42.827160Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Evaluating LLMs When They Do Not Know the Answer: Statistical Evaluation of Mathematical Reasoning via Comparative Signals","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","math.ST","stat.ME","stat.ML","stat.TH"],"primary_cat":"cs.LG","authors_text":"Can Jin, Linjun Zhang, Ruijia Wu, Yang Zhou, Zhixian Zhang, Zihan Dong","submitted_at":"2026-02-03T03:40:01Z","abstract_excerpt":"Evaluating mathematical reasoning in LLMs is constrained by limited benchmark sizes and inherent model stochasticity, yielding high-variance accuracy estimates and unstable rankings across platforms. On difficult problems, an LLM may fail to produce a correct final answer, yet still provide reliable pairwise comparison signals indicating which of two candidate solutions is better. We leverage this observation to design a statistically efficient evaluation framework that combines standard labeled outcomes with pairwise comparison signals obtained by having models judge auxiliary reasoning chain"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.03061","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/2602.03061/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":"2602.03061","created_at":"2026-07-21T01:21:42.827613+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.03061v2","created_at":"2026-07-21T01:21:42.827613+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.03061","created_at":"2026-07-21T01:21:42.827613+00:00"},{"alias_kind":"pith_short_12","alias_value":"X2ICAZ3NMGW5","created_at":"2026-07-21T01:21:42.827613+00:00"},{"alias_kind":"pith_short_16","alias_value":"X2ICAZ3NMGW5BUHC","created_at":"2026-07-21T01:21:42.827613+00:00"},{"alias_kind":"pith_short_8","alias_value":"X2ICAZ3N","created_at":"2026-07-21T01:21:42.827613+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2605.29395","citing_title":"Low Rank for Rank: Uncertainty-Aware Task-Specific LLM Ranking under Sparse Pairwise Comparisons","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2604.02507","citing_title":"Reinforcement Learning from Human Feedback: A Statistical Perspective","ref_index":22,"is_internal_anchor":true},{"citing_arxiv_id":"2604.05460","citing_title":"LLM Evaluation as Tensor Completion: Low Rank Structure and Semiparametric Efficiency","ref_index":7,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/X2ICAZ3NMGW5BUHCYHKPFODSRA","json":"https://pith.science/pith/X2ICAZ3NMGW5BUHCYHKPFODSRA.json","graph_json":"https://pith.science/api/pith-number/X2ICAZ3NMGW5BUHCYHKPFODSRA/graph.json","events_json":"https://pith.science/api/pith-number/X2ICAZ3NMGW5BUHCYHKPFODSRA/events.json","paper":"https://pith.science/paper/X2ICAZ3N"},"agent_actions":{"view_html":"https://pith.science/pith/X2ICAZ3NMGW5BUHCYHKPFODSRA","download_json":"https://pith.science/pith/X2ICAZ3NMGW5BUHCYHKPFODSRA.json","view_paper":"https://pith.science/paper/X2ICAZ3N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.03061&json=true","fetch_graph":"https://pith.science/api/pith-number/X2ICAZ3NMGW5BUHCYHKPFODSRA/graph.json","fetch_events":"https://pith.science/api/pith-number/X2ICAZ3NMGW5BUHCYHKPFODSRA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X2ICAZ3NMGW5BUHCYHKPFODSRA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X2ICAZ3NMGW5BUHCYHKPFODSRA/action/storage_attestation","attest_author":"https://pith.science/pith/X2ICAZ3NMGW5BUHCYHKPFODSRA/action/author_attestation","sign_citation":"https://pith.science/pith/X2ICAZ3NMGW5BUHCYHKPFODSRA/action/citation_signature","submit_replication":"https://pith.science/pith/X2ICAZ3NMGW5BUHCYHKPFODSRA/action/replication_record"}},"created_at":"2026-07-21T01:21:42.827613+00:00","updated_at":"2026-07-21T01:21:42.827613+00:00"}