{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:7XRLVTHDWSVB3BC4XS5XGT4LP7","short_pith_number":"pith:7XRLVTHD","schema_version":"1.0","canonical_sha256":"fde2bacce3b4aa1d845cbcbb734f8b7fe4d17b42988721a762585ceefe448505","source":{"kind":"arxiv","id":"2502.09601","version":1},"attestation_state":"computed","paper":{"title":"CoT-Valve: Length-Compressible Chain-of-Thought Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Gongfan Fang, Guangnian Wan, Runpeng Yu, Xinchao Wang, Xinyin Ma","submitted_at":"2025-02-13T18:52:36Z","abstract_excerpt":"Chain-of-Thought significantly enhances a model's reasoning capability, but it also comes with a considerable increase in inference costs due to long chains. With the observation that the reasoning path can be easily compressed under easy tasks but struggle on hard tasks, we explore the feasibility of elastically controlling the length of reasoning paths with only one model, thereby reducing the inference overhead of reasoning models dynamically based on task difficulty. We introduce a new tuning and inference strategy named CoT-Valve, designed to allow models to generate reasoning chains of v"},"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":"2502.09601","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-02-13T18:52:36Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"fb411e615b727a42f16a8dd55baf87c371ecfdcb2c5051cfbbf8873dde4a7304","abstract_canon_sha256":"1568a7f6bdb67b2a68d3f0aa7670662a2776ad704683e7589a9d0dbfd32a788b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:13:59.909149Z","signature_b64":"X5Q/j3iLKhnUNnybODW+WDSSy1XgIuZY5SXz1nBO10HG3Z5Qk/ARlu7641gCjB/L7nu3uuhAW+7E+/Oy2S4+Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fde2bacce3b4aa1d845cbcbb734f8b7fe4d17b42988721a762585ceefe448505","last_reissued_at":"2026-07-05T10:13:59.908671Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:13:59.908671Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CoT-Valve: Length-Compressible Chain-of-Thought Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Gongfan Fang, Guangnian Wan, Runpeng Yu, Xinchao Wang, Xinyin Ma","submitted_at":"2025-02-13T18:52:36Z","abstract_excerpt":"Chain-of-Thought significantly enhances a model's reasoning capability, but it also comes with a considerable increase in inference costs due to long chains. With the observation that the reasoning path can be easily compressed under easy tasks but struggle on hard tasks, we explore the feasibility of elastically controlling the length of reasoning paths with only one model, thereby reducing the inference overhead of reasoning models dynamically based on task difficulty. We introduce a new tuning and inference strategy named CoT-Valve, designed to allow models to generate reasoning chains of v"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.09601","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/2502.09601/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":"2502.09601","created_at":"2026-07-05T10:13:59.908729+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.09601v1","created_at":"2026-07-05T10:13:59.908729+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.09601","created_at":"2026-07-05T10:13:59.908729+00:00"},{"alias_kind":"pith_short_12","alias_value":"7XRLVTHDWSVB","created_at":"2026-07-05T10:13:59.908729+00:00"},{"alias_kind":"pith_short_16","alias_value":"7XRLVTHDWSVB3BC4","created_at":"2026-07-05T10:13:59.908729+00:00"},{"alias_kind":"pith_short_8","alias_value":"7XRLVTHD","created_at":"2026-07-05T10:13:59.908729+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01532","citing_title":"Rethinking the Role of Positional Encoding: Sliding-Window Transformers without PE Remain Turing Complete","ref_index":88,"is_internal_anchor":false},{"citing_arxiv_id":"2602.07832","citing_title":"rePIRL: Learn PRM with Inverse RL for LLM Reasoning","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2509.05489","citing_title":"Self-Aligned Reward: Towards Effective and Efficient Reasoners","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13165","citing_title":"STOP: Structured On-Policy Pruning of Long-Form Reasoning in Low-Data Regimes","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2503.16419","citing_title":"Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models","ref_index":130,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06165","citing_title":"Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14847","citing_title":"TrigReason: Trigger-Based Collaboration between Small and Large Reasoning Models","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7XRLVTHDWSVB3BC4XS5XGT4LP7","json":"https://pith.science/pith/7XRLVTHDWSVB3BC4XS5XGT4LP7.json","graph_json":"https://pith.science/api/pith-number/7XRLVTHDWSVB3BC4XS5XGT4LP7/graph.json","events_json":"https://pith.science/api/pith-number/7XRLVTHDWSVB3BC4XS5XGT4LP7/events.json","paper":"https://pith.science/paper/7XRLVTHD"},"agent_actions":{"view_html":"https://pith.science/pith/7XRLVTHDWSVB3BC4XS5XGT4LP7","download_json":"https://pith.science/pith/7XRLVTHDWSVB3BC4XS5XGT4LP7.json","view_paper":"https://pith.science/paper/7XRLVTHD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.09601&json=true","fetch_graph":"https://pith.science/api/pith-number/7XRLVTHDWSVB3BC4XS5XGT4LP7/graph.json","fetch_events":"https://pith.science/api/pith-number/7XRLVTHDWSVB3BC4XS5XGT4LP7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7XRLVTHDWSVB3BC4XS5XGT4LP7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7XRLVTHDWSVB3BC4XS5XGT4LP7/action/storage_attestation","attest_author":"https://pith.science/pith/7XRLVTHDWSVB3BC4XS5XGT4LP7/action/author_attestation","sign_citation":"https://pith.science/pith/7XRLVTHDWSVB3BC4XS5XGT4LP7/action/citation_signature","submit_replication":"https://pith.science/pith/7XRLVTHDWSVB3BC4XS5XGT4LP7/action/replication_record"}},"created_at":"2026-07-05T10:13:59.908729+00:00","updated_at":"2026-07-05T10:13:59.908729+00:00"}