{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:63RDQTHZ3GG6GFHGVBSLATVQMI","short_pith_number":"pith:63RDQTHZ","schema_version":"1.0","canonical_sha256":"f6e2384cf9d98de314e6a864b04eb0622b08b5451158aead79099fec2b1695bd","source":{"kind":"arxiv","id":"2506.03295","version":2},"attestation_state":"computed","paper":{"title":"Unleashing the Reasoning Potential of Pre-trained LLMs by Critique Fine-Tuning on One Problem","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Kai Zou, Lijun Wu, Ping Nie, Wenhu Chen, Yubo Wang","submitted_at":"2025-06-03T18:35:52Z","abstract_excerpt":"We have witnessed that strong LLMs like Qwen-Math, MiMo, and Phi-4 possess immense reasoning potential inherited from the pre-training stage. With reinforcement learning (RL), these models can improve dramatically on reasoning tasks. Recent studies have shown that even RL on a single problem can unleash these models' reasoning capabilities. However, RL is not only expensive but also unstable. Even one-shot RL requires hundreds of GPU hours. This raises a critical question: Is there a more efficient way to unleash the reasoning potential of these powerful base LLMs? In this work, we demonstrate"},"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":"2506.03295","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-03T18:35:52Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"f2a55410e31bf4fe532c797624ffec6c5a1206067436a51d5cd591b1350cd7a9","abstract_canon_sha256":"fb5669e5a22b4b0bd49910573156d1aec18cad25a75de8660e792a43944cf505"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:16:09.798252Z","signature_b64":"hCVhaCArrqUdIDTU+cvN/TUK1vCWjFjx6vzr/NABf3l5S7QUNk0zCZYP3p8fJribwrLEXB9CV+SWKltIk4i3BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f6e2384cf9d98de314e6a864b04eb0622b08b5451158aead79099fec2b1695bd","last_reissued_at":"2026-07-05T11:16:09.797663Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:16:09.797663Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unleashing the Reasoning Potential of Pre-trained LLMs by Critique Fine-Tuning on One Problem","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Kai Zou, Lijun Wu, Ping Nie, Wenhu Chen, Yubo Wang","submitted_at":"2025-06-03T18:35:52Z","abstract_excerpt":"We have witnessed that strong LLMs like Qwen-Math, MiMo, and Phi-4 possess immense reasoning potential inherited from the pre-training stage. With reinforcement learning (RL), these models can improve dramatically on reasoning tasks. Recent studies have shown that even RL on a single problem can unleash these models' reasoning capabilities. However, RL is not only expensive but also unstable. Even one-shot RL requires hundreds of GPU hours. This raises a critical question: Is there a more efficient way to unleash the reasoning potential of these powerful base LLMs? In this work, we demonstrate"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.03295","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/2506.03295/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":"2506.03295","created_at":"2026-07-05T11:16:09.797729+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.03295v2","created_at":"2026-07-05T11:16:09.797729+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.03295","created_at":"2026-07-05T11:16:09.797729+00:00"},{"alias_kind":"pith_short_12","alias_value":"63RDQTHZ3GG6","created_at":"2026-07-05T11:16:09.797729+00:00"},{"alias_kind":"pith_short_16","alias_value":"63RDQTHZ3GG6GFHG","created_at":"2026-07-05T11:16:09.797729+00:00"},{"alias_kind":"pith_short_8","alias_value":"63RDQTHZ","created_at":"2026-07-05T11:16:09.797729+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.17928","citing_title":"HEALing Entropy Collapse: Enhancing Exploration in Few-Shot RLVR via Hybrid-Domain Entropy Dynamics Alignment","ref_index":35,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/63RDQTHZ3GG6GFHGVBSLATVQMI","json":"https://pith.science/pith/63RDQTHZ3GG6GFHGVBSLATVQMI.json","graph_json":"https://pith.science/api/pith-number/63RDQTHZ3GG6GFHGVBSLATVQMI/graph.json","events_json":"https://pith.science/api/pith-number/63RDQTHZ3GG6GFHGVBSLATVQMI/events.json","paper":"https://pith.science/paper/63RDQTHZ"},"agent_actions":{"view_html":"https://pith.science/pith/63RDQTHZ3GG6GFHGVBSLATVQMI","download_json":"https://pith.science/pith/63RDQTHZ3GG6GFHGVBSLATVQMI.json","view_paper":"https://pith.science/paper/63RDQTHZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.03295&json=true","fetch_graph":"https://pith.science/api/pith-number/63RDQTHZ3GG6GFHGVBSLATVQMI/graph.json","fetch_events":"https://pith.science/api/pith-number/63RDQTHZ3GG6GFHGVBSLATVQMI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/63RDQTHZ3GG6GFHGVBSLATVQMI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/63RDQTHZ3GG6GFHGVBSLATVQMI/action/storage_attestation","attest_author":"https://pith.science/pith/63RDQTHZ3GG6GFHGVBSLATVQMI/action/author_attestation","sign_citation":"https://pith.science/pith/63RDQTHZ3GG6GFHGVBSLATVQMI/action/citation_signature","submit_replication":"https://pith.science/pith/63RDQTHZ3GG6GFHGVBSLATVQMI/action/replication_record"}},"created_at":"2026-07-05T11:16:09.797729+00:00","updated_at":"2026-07-05T11:16:09.797729+00:00"}