{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FIHGPDF7BULGDLCFZZLFEGAHD2","short_pith_number":"pith:FIHGPDF7","schema_version":"1.0","canonical_sha256":"2a0e678cbf0d1661ac45ce565218071eb339ef8401719792762fe4592b42e4bd","source":{"kind":"arxiv","id":"2505.14391","version":1},"attestation_state":"computed","paper":{"title":"Beyond the First Error: Process Reward Models for Reflective Mathematical Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chenghua He, Daxin Jiang, Linjing Li, Qiyue Yin, Shihong Deng, Xiaowen Shi, Zhaohui Yang","submitted_at":"2025-05-20T14:12:05Z","abstract_excerpt":"Many studies focus on data annotation techniques for training effective PRMs. However, current methods encounter a significant issue when applied to long CoT reasoning processes: they tend to focus solely on the first incorrect step and all preceding steps, assuming that all subsequent steps are incorrect. These methods overlook the unique self-correction and reflection mechanisms inherent in long CoT, where correct reasoning steps may still occur after initial reasoning mistakes. To address this issue, we propose a novel data annotation method for PRMs specifically designed to score the long "},"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":"2505.14391","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-05-20T14:12:05Z","cross_cats_sorted":[],"title_canon_sha256":"46ab7c6d2d0d1649b0d9d0d78a6abb708871ba118a8176bb1e4ca0dcffbff8fe","abstract_canon_sha256":"12a6687911ddf419c05e522e0096e57c78fd7b2b5b9c2d4888b4deac7af916ed"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:06:00.770174Z","signature_b64":"d6SeNAwwq7O8H6IBpoecCXmOL76VMA1PdhTD/Xrk+wXgPZV9ADWpW0Gwx9a0+DK3BjgW2v5JD4Y4ONnCRhTFDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2a0e678cbf0d1661ac45ce565218071eb339ef8401719792762fe4592b42e4bd","last_reissued_at":"2026-07-05T11:06:00.769747Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:06:00.769747Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond the First Error: Process Reward Models for Reflective Mathematical Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chenghua He, Daxin Jiang, Linjing Li, Qiyue Yin, Shihong Deng, Xiaowen Shi, Zhaohui Yang","submitted_at":"2025-05-20T14:12:05Z","abstract_excerpt":"Many studies focus on data annotation techniques for training effective PRMs. However, current methods encounter a significant issue when applied to long CoT reasoning processes: they tend to focus solely on the first incorrect step and all preceding steps, assuming that all subsequent steps are incorrect. These methods overlook the unique self-correction and reflection mechanisms inherent in long CoT, where correct reasoning steps may still occur after initial reasoning mistakes. To address this issue, we propose a novel data annotation method for PRMs specifically designed to score the long "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.14391","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/2505.14391/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":"2505.14391","created_at":"2026-07-05T11:06:00.769803+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.14391v1","created_at":"2026-07-05T11:06:00.769803+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.14391","created_at":"2026-07-05T11:06:00.769803+00:00"},{"alias_kind":"pith_short_12","alias_value":"FIHGPDF7BULG","created_at":"2026-07-05T11:06:00.769803+00:00"},{"alias_kind":"pith_short_16","alias_value":"FIHGPDF7BULGDLCF","created_at":"2026-07-05T11:06:00.769803+00:00"},{"alias_kind":"pith_short_8","alias_value":"FIHGPDF7","created_at":"2026-07-05T11:06:00.769803+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.02341","citing_title":"LLM Reasoning with Process Rewards for Outcome-Guided Steps","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13467","citing_title":"PDCR: Perception-Decomposed Confidence Reward for Vision-Language Reasoning","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04064","citing_title":"Improving Medical VQA through Trajectory-Aware Process Supervision","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05226","citing_title":"Internalizing Outcome Supervision into Process Supervision: A New Paradigm for Reinforcement Learning for Reasoning","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FIHGPDF7BULGDLCFZZLFEGAHD2","json":"https://pith.science/pith/FIHGPDF7BULGDLCFZZLFEGAHD2.json","graph_json":"https://pith.science/api/pith-number/FIHGPDF7BULGDLCFZZLFEGAHD2/graph.json","events_json":"https://pith.science/api/pith-number/FIHGPDF7BULGDLCFZZLFEGAHD2/events.json","paper":"https://pith.science/paper/FIHGPDF7"},"agent_actions":{"view_html":"https://pith.science/pith/FIHGPDF7BULGDLCFZZLFEGAHD2","download_json":"https://pith.science/pith/FIHGPDF7BULGDLCFZZLFEGAHD2.json","view_paper":"https://pith.science/paper/FIHGPDF7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.14391&json=true","fetch_graph":"https://pith.science/api/pith-number/FIHGPDF7BULGDLCFZZLFEGAHD2/graph.json","fetch_events":"https://pith.science/api/pith-number/FIHGPDF7BULGDLCFZZLFEGAHD2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FIHGPDF7BULGDLCFZZLFEGAHD2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FIHGPDF7BULGDLCFZZLFEGAHD2/action/storage_attestation","attest_author":"https://pith.science/pith/FIHGPDF7BULGDLCFZZLFEGAHD2/action/author_attestation","sign_citation":"https://pith.science/pith/FIHGPDF7BULGDLCFZZLFEGAHD2/action/citation_signature","submit_replication":"https://pith.science/pith/FIHGPDF7BULGDLCFZZLFEGAHD2/action/replication_record"}},"created_at":"2026-07-05T11:06:00.769803+00:00","updated_at":"2026-07-05T11:06:00.769803+00:00"}