{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2ACTI7WIN35LY2WUGWSCWKP4ND","short_pith_number":"pith:2ACTI7WI","schema_version":"1.0","canonical_sha256":"d005347ec86efabc6ad435a42b29fc68f0974c16244f24cf36219d7f6a3b4a65","source":{"kind":"arxiv","id":"2503.01461","version":2},"attestation_state":"computed","paper":{"title":"Marco-o1 v2: Towards Widening The Distillation Bottleneck for Reasoning Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Bo Zeng, Chenyang Lyu, Hao Wang, Huifeng Yin, Kaifu Zhang, Liangying Shao, Longyue Wang, Minghao Wu, Tianqi Shi, Weihua Luo, Xuanfan Ni, Yu Zhao","submitted_at":"2025-03-03T12:17:36Z","abstract_excerpt":"Large Reasoning Models(LRMs) such as OpenAI o1 and DeepSeek-R1 have shown remarkable reasoning capabilities by scaling test-time compute and generating long Chain-of-Thought(CoT). Distillation--post-training on LRMs-generated data--is a straightforward yet effective method to enhance the reasoning abilities of smaller models, but faces a critical bottleneck: we found that distilled long CoT data poses learning difficulty for small models and leads to the inheritance of biases (i.e. over-thinking) when using Supervised Fine-tuning (SFT) and Reinforcement Learning (RL) methods. To alleviate this"},"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":"2503.01461","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-03T12:17:36Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"ee6dc51cc20c2c00dbe34523f13e6f0e52e9ef042995d7cc27ac5f085b31f525","abstract_canon_sha256":"808907210a5ca8ee2e9f2301d317452c223913818ff22934d449de1690a5463a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:21.817970Z","signature_b64":"FqXtgUiTeXnTNGb69SQkMCZDYc4tc8ORF8116heLoWuAcEm4J5s1X0ZE6EOmUF+6DT1NkQ6CNVTTkz0MiEgUBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d005347ec86efabc6ad435a42b29fc68f0974c16244f24cf36219d7f6a3b4a65","last_reissued_at":"2026-07-05T11:13:21.817530Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:21.817530Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Marco-o1 v2: Towards Widening The Distillation Bottleneck for Reasoning Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Bo Zeng, Chenyang Lyu, Hao Wang, Huifeng Yin, Kaifu Zhang, Liangying Shao, Longyue Wang, Minghao Wu, Tianqi Shi, Weihua Luo, Xuanfan Ni, Yu Zhao","submitted_at":"2025-03-03T12:17:36Z","abstract_excerpt":"Large Reasoning Models(LRMs) such as OpenAI o1 and DeepSeek-R1 have shown remarkable reasoning capabilities by scaling test-time compute and generating long Chain-of-Thought(CoT). Distillation--post-training on LRMs-generated data--is a straightforward yet effective method to enhance the reasoning abilities of smaller models, but faces a critical bottleneck: we found that distilled long CoT data poses learning difficulty for small models and leads to the inheritance of biases (i.e. over-thinking) when using Supervised Fine-tuning (SFT) and Reinforcement Learning (RL) methods. To alleviate this"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.01461","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/2503.01461/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":"2503.01461","created_at":"2026-07-05T11:13:21.817586+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.01461v2","created_at":"2026-07-05T11:13:21.817586+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.01461","created_at":"2026-07-05T11:13:21.817586+00:00"},{"alias_kind":"pith_short_12","alias_value":"2ACTI7WIN35L","created_at":"2026-07-05T11:13:21.817586+00:00"},{"alias_kind":"pith_short_16","alias_value":"2ACTI7WIN35LY2WU","created_at":"2026-07-05T11:13:21.817586+00:00"},{"alias_kind":"pith_short_8","alias_value":"2ACTI7WI","created_at":"2026-07-05T11:13:21.817586+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/2ACTI7WIN35LY2WUGWSCWKP4ND","json":"https://pith.science/pith/2ACTI7WIN35LY2WUGWSCWKP4ND.json","graph_json":"https://pith.science/api/pith-number/2ACTI7WIN35LY2WUGWSCWKP4ND/graph.json","events_json":"https://pith.science/api/pith-number/2ACTI7WIN35LY2WUGWSCWKP4ND/events.json","paper":"https://pith.science/paper/2ACTI7WI"},"agent_actions":{"view_html":"https://pith.science/pith/2ACTI7WIN35LY2WUGWSCWKP4ND","download_json":"https://pith.science/pith/2ACTI7WIN35LY2WUGWSCWKP4ND.json","view_paper":"https://pith.science/paper/2ACTI7WI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.01461&json=true","fetch_graph":"https://pith.science/api/pith-number/2ACTI7WIN35LY2WUGWSCWKP4ND/graph.json","fetch_events":"https://pith.science/api/pith-number/2ACTI7WIN35LY2WUGWSCWKP4ND/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2ACTI7WIN35LY2WUGWSCWKP4ND/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2ACTI7WIN35LY2WUGWSCWKP4ND/action/storage_attestation","attest_author":"https://pith.science/pith/2ACTI7WIN35LY2WUGWSCWKP4ND/action/author_attestation","sign_citation":"https://pith.science/pith/2ACTI7WIN35LY2WUGWSCWKP4ND/action/citation_signature","submit_replication":"https://pith.science/pith/2ACTI7WIN35LY2WUGWSCWKP4ND/action/replication_record"}},"created_at":"2026-07-05T11:13:21.817586+00:00","updated_at":"2026-07-05T11:13:21.817586+00:00"}