{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BIGRIRYTURQNWR3X3SCVO52MCJ","short_pith_number":"pith:BIGRIRYT","schema_version":"1.0","canonical_sha256":"0a0d144713a460db4777dc8557774c1260cdb8035fa20d393f5895cfa8b83415","source":{"kind":"arxiv","id":"2503.04697","version":2},"attestation_state":"computed","paper":{"title":"L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Pranjal Aggarwal, Sean Welleck","submitted_at":"2025-03-06T18:43:29Z","abstract_excerpt":"Reasoning language models have shown an uncanny ability to improve performance at test-time by ``thinking longer''-that is, by generating longer chain-of-thought sequences and hence using more compute. However, the length of their chain-of-thought reasoning is not controllable, making it impossible to allocate test-time compute to achieve a desired level of performance. We introduce Length Controlled Policy Optimization (LCPO), a simple reinforcement learning method that optimizes for accuracy and adherence to user-specified length constraints. We use LCPO to train L1, a reasoning language mod"},"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":"2503.04697","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-03-06T18:43:29Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"6a2608bda5390a148c845d3ccc314878e6e45229cca6f039c066cdae4229a6b0","abstract_canon_sha256":"cd2f52aefc144d041b237101c846baebaa58a1db7f04a77659b8137b10179faf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:13:29.491814Z","signature_b64":"Z6Q8pnxQgV3YxWWzAm1qd8SS7mxCX7P038Ertk/CuFUDaQdWVtOhZILqAPnQv2k5kn6+dOyTvZ06bHIlfgcqBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0a0d144713a460db4777dc8557774c1260cdb8035fa20d393f5895cfa8b83415","last_reissued_at":"2026-05-18T00:13:29.491093Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:13:29.491093Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Pranjal Aggarwal, Sean Welleck","submitted_at":"2025-03-06T18:43:29Z","abstract_excerpt":"Reasoning language models have shown an uncanny ability to improve performance at test-time by ``thinking longer''-that is, by generating longer chain-of-thought sequences and hence using more compute. However, the length of their chain-of-thought reasoning is not controllable, making it impossible to allocate test-time compute to achieve a desired level of performance. We introduce Length Controlled Policy Optimization (LCPO), a simple reinforcement learning method that optimizes for accuracy and adherence to user-specified length constraints. We use LCPO to train L1, a reasoning language mod"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.04697","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":""},"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":"2503.04697","created_at":"2026-05-18T00:13:29.491187+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.04697v2","created_at":"2026-05-18T00:13:29.491187+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.04697","created_at":"2026-05-18T00:13:29.491187+00:00"},{"alias_kind":"pith_short_12","alias_value":"BIGRIRYTURQN","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"BIGRIRYTURQNWR3X","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"BIGRIRYT","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":25,"internal_anchor_count":25,"sample":[{"citing_arxiv_id":"2504.12501","citing_title":"Reinforcement Learning from Human Feedback","ref_index":192,"is_internal_anchor":true},{"citing_arxiv_id":"2605.22138","citing_title":"Efficient Agentic Reasoning Through Self-Regulated Simulative Planning","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2605.22211","citing_title":"CLORE: Content-Level Optimization for Reasoning Efficiency","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2605.21463","citing_title":"Mem-$\\pi$: Adaptive Memory through Learning When and What to Generate","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2605.19358","citing_title":"Taming the Thinker: Conditional Entropy Shaping for Adaptive LLM Reasoning","ref_index":17,"is_internal_anchor":true},{"citing_arxiv_id":"2507.12549","citing_title":"The Serial Scaling Hypothesis","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2508.10164","citing_title":"Pruning Long Chain-of-Thought of Large Reasoning Models via Small-Scale Preference Optimization","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2509.05489","citing_title":"Self-Aligned Reward: Towards Effective and Efficient Reasoners","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2509.08827","citing_title":"A Survey of Reinforcement Learning for Large Reasoning Models","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2601.05300","citing_title":"TIME: Temporally Intelligent Meta-reasoning Engine for Context-Triggered Explicit Reasoning","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2503.21620","citing_title":"UI-R1: Enhancing Efficient Action Prediction of GUI Agents by Reinforcement Learning","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2603.05433","citing_title":"CRISP: Compressed Reasoning via Iterative Self-Policy Distillation","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2601.05242","citing_title":"GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2603.22267","citing_title":"TiCo: Time-Controllable Spoken Dialogue Model","ref_index":45,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13165","citing_title":"STOP: Structured On-Policy Pruning of Long-Form Reasoning in Low-Data Regimes","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13641","citing_title":"Multi-Objective and Mixed-Reward Reinforcement Learning via Reward-Decorrelated Policy Optimization","ref_index":5,"is_internal_anchor":true},{"citing_arxiv_id":"2503.16419","citing_title":"Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2604.04120","citing_title":"Shorter, but Still Trustworthy? An Empirical Study of Chain-of-Thought Compression","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11625","citing_title":"Nice Fold or Hero Call: Learning Budget-Efficient Thinking for Adaptive Reasoning","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2605.08221","citing_title":"NoisyCoconut: Counterfactual Consensus via Latent Space Reasoning","ref_index":95,"is_internal_anchor":true},{"citing_arxiv_id":"2503.09567","citing_title":"Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.08776","citing_title":"Reasoning Compression with Mixed-Policy Distillation","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09806","citing_title":"LEAD: Length-Efficient Adaptive and Dynamic Reasoning for Large Language Models","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06165","citing_title":"Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost","ref_index":66,"is_internal_anchor":true},{"citing_arxiv_id":"2604.10034","citing_title":"AI Achieves a Perfect LSAT Score","ref_index":21,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BIGRIRYTURQNWR3X3SCVO52MCJ","json":"https://pith.science/pith/BIGRIRYTURQNWR3X3SCVO52MCJ.json","graph_json":"https://pith.science/api/pith-number/BIGRIRYTURQNWR3X3SCVO52MCJ/graph.json","events_json":"https://pith.science/api/pith-number/BIGRIRYTURQNWR3X3SCVO52MCJ/events.json","paper":"https://pith.science/paper/BIGRIRYT"},"agent_actions":{"view_html":"https://pith.science/pith/BIGRIRYTURQNWR3X3SCVO52MCJ","download_json":"https://pith.science/pith/BIGRIRYTURQNWR3X3SCVO52MCJ.json","view_paper":"https://pith.science/paper/BIGRIRYT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.04697&json=true","fetch_graph":"https://pith.science/api/pith-number/BIGRIRYTURQNWR3X3SCVO52MCJ/graph.json","fetch_events":"https://pith.science/api/pith-number/BIGRIRYTURQNWR3X3SCVO52MCJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BIGRIRYTURQNWR3X3SCVO52MCJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BIGRIRYTURQNWR3X3SCVO52MCJ/action/storage_attestation","attest_author":"https://pith.science/pith/BIGRIRYTURQNWR3X3SCVO52MCJ/action/author_attestation","sign_citation":"https://pith.science/pith/BIGRIRYTURQNWR3X3SCVO52MCJ/action/citation_signature","submit_replication":"https://pith.science/pith/BIGRIRYTURQNWR3X3SCVO52MCJ/action/replication_record"}},"created_at":"2026-05-18T00:13:29.491187+00:00","updated_at":"2026-05-18T00:13:29.491187+00:00"}