{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:2JFGQVOQ57FIZ26FVI7QSUD5VU","short_pith_number":"pith:2JFGQVOQ","schema_version":"1.0","canonical_sha256":"d24a6855d0efca8cebc5aa3f09507dad0132faad6d1cebc16d97cdcf432d565d","source":{"kind":"arxiv","id":"2404.10346","version":4},"attestation_state":"computed","paper":{"title":"Self-Explore: Enhancing Mathematical Reasoning in Language Models with Fine-grained Rewards","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Doyoung Kim, Hyeonbin Hwang, Minjoon Seo, Seonghyeon Ye, Seungone Kim","submitted_at":"2024-04-16T07:30:11Z","abstract_excerpt":"Training on large amounts of rationales (i.e., CoT Fine-tuning) is effective at improving the reasoning capabilities of large language models (LLMs). However, acquiring human-authored rationales or augmenting rationales from proprietary models is costly and not scalable. In this paper, we study the problem of whether LLMs could self-improve their reasoning capabilities. To this end, we propose Self-Explore, where the LLM is tasked to explore the first wrong step (i.e., the first pit) within the rationale and use such signals as fine-grained rewards for further improvement. On the GSM8K and MAT"},"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":"2404.10346","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-04-16T07:30:11Z","cross_cats_sorted":[],"title_canon_sha256":"006cce95b1a70372739183657286943d5d67d2e5ce373d9f02a234960f35879b","abstract_canon_sha256":"0b7f1adf1294302053617fcfd54ef67c6054b121f18185b226f6f01eb0e8b101"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:15:09.850955Z","signature_b64":"qdzeHMv0InF334irj7JnyHmBR5bWhiirUorOl4d9XoJmjjA43OPnhXMvlUH0A0WvJ26ZhU8yrvbYDgn6j9P2Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d24a6855d0efca8cebc5aa3f09507dad0132faad6d1cebc16d97cdcf432d565d","last_reissued_at":"2026-07-05T09:15:09.850491Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:15:09.850491Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Self-Explore: Enhancing Mathematical Reasoning in Language Models with Fine-grained Rewards","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Doyoung Kim, Hyeonbin Hwang, Minjoon Seo, Seonghyeon Ye, Seungone Kim","submitted_at":"2024-04-16T07:30:11Z","abstract_excerpt":"Training on large amounts of rationales (i.e., CoT Fine-tuning) is effective at improving the reasoning capabilities of large language models (LLMs). However, acquiring human-authored rationales or augmenting rationales from proprietary models is costly and not scalable. In this paper, we study the problem of whether LLMs could self-improve their reasoning capabilities. To this end, we propose Self-Explore, where the LLM is tasked to explore the first wrong step (i.e., the first pit) within the rationale and use such signals as fine-grained rewards for further improvement. On the GSM8K and MAT"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.10346","kind":"arxiv","version":4},"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/2404.10346/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":"2404.10346","created_at":"2026-07-05T09:15:09.850551+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.10346v4","created_at":"2026-07-05T09:15:09.850551+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.10346","created_at":"2026-07-05T09:15:09.850551+00:00"},{"alias_kind":"pith_short_12","alias_value":"2JFGQVOQ57FI","created_at":"2026-07-05T09:15:09.850551+00:00"},{"alias_kind":"pith_short_16","alias_value":"2JFGQVOQ57FIZ26F","created_at":"2026-07-05T09:15:09.850551+00:00"},{"alias_kind":"pith_short_8","alias_value":"2JFGQVOQ","created_at":"2026-07-05T09:15:09.850551+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2410.08146","citing_title":"Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoning","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2501.09686","citing_title":"Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models","ref_index":56,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2JFGQVOQ57FIZ26FVI7QSUD5VU","json":"https://pith.science/pith/2JFGQVOQ57FIZ26FVI7QSUD5VU.json","graph_json":"https://pith.science/api/pith-number/2JFGQVOQ57FIZ26FVI7QSUD5VU/graph.json","events_json":"https://pith.science/api/pith-number/2JFGQVOQ57FIZ26FVI7QSUD5VU/events.json","paper":"https://pith.science/paper/2JFGQVOQ"},"agent_actions":{"view_html":"https://pith.science/pith/2JFGQVOQ57FIZ26FVI7QSUD5VU","download_json":"https://pith.science/pith/2JFGQVOQ57FIZ26FVI7QSUD5VU.json","view_paper":"https://pith.science/paper/2JFGQVOQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.10346&json=true","fetch_graph":"https://pith.science/api/pith-number/2JFGQVOQ57FIZ26FVI7QSUD5VU/graph.json","fetch_events":"https://pith.science/api/pith-number/2JFGQVOQ57FIZ26FVI7QSUD5VU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2JFGQVOQ57FIZ26FVI7QSUD5VU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2JFGQVOQ57FIZ26FVI7QSUD5VU/action/storage_attestation","attest_author":"https://pith.science/pith/2JFGQVOQ57FIZ26FVI7QSUD5VU/action/author_attestation","sign_citation":"https://pith.science/pith/2JFGQVOQ57FIZ26FVI7QSUD5VU/action/citation_signature","submit_replication":"https://pith.science/pith/2JFGQVOQ57FIZ26FVI7QSUD5VU/action/replication_record"}},"created_at":"2026-07-05T09:15:09.850551+00:00","updated_at":"2026-07-05T09:15:09.850551+00:00"}