{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:Y32X4CST5B762NCJSGWYRV4IXZ","short_pith_number":"pith:Y32X4CST","schema_version":"1.0","canonical_sha256":"c6f57e0a53e87fed344991ad88d788be5ac982178646b097ab389dec033bd0db","source":{"kind":"arxiv","id":"2506.17533","version":1},"attestation_state":"computed","paper":{"title":"DuaShepherd: Integrating Stepwise Correctness and Potential Rewards for Mathematical Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Cheng Niu, Hanning Zhang, Juntong Song, Tong Zhang, Yuanhao Wu","submitted_at":"2025-06-21T01:11:01Z","abstract_excerpt":"In this paper, we propose DuaShepherd, a novel reward modeling framework that integrates two complementary reward signals, correctness and potential, to enhance the mathematical reasoning capabilities of Large Language Models (LLMs). While correctness-based signals emphasize identification of stepwise errors, potential-based signals focus on the likelihood of reaching the correct final answer. We developed an automated pipeline for constructing large-scale reward modeling dataset with both signals. A unified, multi-head architecture was explored to train the two reward models in a multi-task s"},"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":"2506.17533","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-21T01:11:01Z","cross_cats_sorted":[],"title_canon_sha256":"f1383273af49187ebc70455ebeaac5a7e5a0433401a08e1653e49f748d705332","abstract_canon_sha256":"ec8eee7f45a8935dba76129bae2739b19a80ff3c25164ad25aa68c30534c0c3b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:25:19.042247Z","signature_b64":"3EkuGX1QMe698LhzrkdqEPEpxstXiO/iIBH4k5Pq49jJrzAvgNG4gvH2nT6Sr4HIrnD/nnzI/h6wG8phIVYpAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c6f57e0a53e87fed344991ad88d788be5ac982178646b097ab389dec033bd0db","last_reissued_at":"2026-07-05T11:25:19.041778Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:25:19.041778Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DuaShepherd: Integrating Stepwise Correctness and Potential Rewards for Mathematical Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Cheng Niu, Hanning Zhang, Juntong Song, Tong Zhang, Yuanhao Wu","submitted_at":"2025-06-21T01:11:01Z","abstract_excerpt":"In this paper, we propose DuaShepherd, a novel reward modeling framework that integrates two complementary reward signals, correctness and potential, to enhance the mathematical reasoning capabilities of Large Language Models (LLMs). While correctness-based signals emphasize identification of stepwise errors, potential-based signals focus on the likelihood of reaching the correct final answer. We developed an automated pipeline for constructing large-scale reward modeling dataset with both signals. A unified, multi-head architecture was explored to train the two reward models in a multi-task s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.17533","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/2506.17533/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":"2506.17533","created_at":"2026-07-05T11:25:19.041835+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.17533v1","created_at":"2026-07-05T11:25:19.041835+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.17533","created_at":"2026-07-05T11:25:19.041835+00:00"},{"alias_kind":"pith_short_12","alias_value":"Y32X4CST5B76","created_at":"2026-07-05T11:25:19.041835+00:00"},{"alias_kind":"pith_short_16","alias_value":"Y32X4CST5B762NCJ","created_at":"2026-07-05T11:25:19.041835+00:00"},{"alias_kind":"pith_short_8","alias_value":"Y32X4CST","created_at":"2026-07-05T11:25:19.041835+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Y32X4CST5B762NCJSGWYRV4IXZ","json":"https://pith.science/pith/Y32X4CST5B762NCJSGWYRV4IXZ.json","graph_json":"https://pith.science/api/pith-number/Y32X4CST5B762NCJSGWYRV4IXZ/graph.json","events_json":"https://pith.science/api/pith-number/Y32X4CST5B762NCJSGWYRV4IXZ/events.json","paper":"https://pith.science/paper/Y32X4CST"},"agent_actions":{"view_html":"https://pith.science/pith/Y32X4CST5B762NCJSGWYRV4IXZ","download_json":"https://pith.science/pith/Y32X4CST5B762NCJSGWYRV4IXZ.json","view_paper":"https://pith.science/paper/Y32X4CST","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.17533&json=true","fetch_graph":"https://pith.science/api/pith-number/Y32X4CST5B762NCJSGWYRV4IXZ/graph.json","fetch_events":"https://pith.science/api/pith-number/Y32X4CST5B762NCJSGWYRV4IXZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Y32X4CST5B762NCJSGWYRV4IXZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Y32X4CST5B762NCJSGWYRV4IXZ/action/storage_attestation","attest_author":"https://pith.science/pith/Y32X4CST5B762NCJSGWYRV4IXZ/action/author_attestation","sign_citation":"https://pith.science/pith/Y32X4CST5B762NCJSGWYRV4IXZ/action/citation_signature","submit_replication":"https://pith.science/pith/Y32X4CST5B762NCJSGWYRV4IXZ/action/replication_record"}},"created_at":"2026-07-05T11:25:19.041835+00:00","updated_at":"2026-07-05T11:25:19.041835+00:00"}