{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:N6PCWXV7OGXO4EIGFTKTZJUEW7","short_pith_number":"pith:N6PCWXV7","schema_version":"1.0","canonical_sha256":"6f9e2b5ebf71aeee11062cd53ca684b7d2d491151074742f2732adfe93bafc1b","source":{"kind":"arxiv","id":"2508.08940","version":1},"attestation_state":"computed","paper":{"title":"Train Long, Think Short: Curriculum Learning for Efficient Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Abed Hammoud, Bernard Ghanem, Elie Bou-Zeid, Hasan Abed Al Kader Hammoud, Kumail Alhamoud, Marzyeh Ghassemi","submitted_at":"2025-08-12T13:48:03Z","abstract_excerpt":"Recent work on enhancing the reasoning abilities of large language models (LLMs) has introduced explicit length control as a means of constraining computational cost while preserving accuracy. However, existing approaches rely on fixed-length training budgets, which do not take advantage of the natural progression from exploration to compression during learning. In this work, we propose a curriculum learning strategy for length-controlled reasoning using Group Relative Policy Optimization (GRPO). Our method starts with generous token budgets and gradually tightens them over training, encouragi"},"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":"2508.08940","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-12T13:48:03Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"7bb3ef54eba45e2c9056885940e4459d38de2e8a3b8f92e6e709ca1a2b7d97af","abstract_canon_sha256":"a11110ddd7815b5c9948ada730657a53b58f2d058f89c24f56e1941938edf5f5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:52:37.068478Z","signature_b64":"VtxYuaLWPzMiBaOkFWuZmjCK08FdKrXA9XIXrKc1DhMZ/FyFueIznp1CG3vRT39pj7tzWjdovr8Q1Zg8U02wAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6f9e2b5ebf71aeee11062cd53ca684b7d2d491151074742f2732adfe93bafc1b","last_reissued_at":"2026-07-05T11:52:37.067783Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:52:37.067783Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Train Long, Think Short: Curriculum Learning for Efficient Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Abed Hammoud, Bernard Ghanem, Elie Bou-Zeid, Hasan Abed Al Kader Hammoud, Kumail Alhamoud, Marzyeh Ghassemi","submitted_at":"2025-08-12T13:48:03Z","abstract_excerpt":"Recent work on enhancing the reasoning abilities of large language models (LLMs) has introduced explicit length control as a means of constraining computational cost while preserving accuracy. However, existing approaches rely on fixed-length training budgets, which do not take advantage of the natural progression from exploration to compression during learning. In this work, we propose a curriculum learning strategy for length-controlled reasoning using Group Relative Policy Optimization (GRPO). Our method starts with generous token budgets and gradually tightens them over training, encouragi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.08940","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/2508.08940/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":"2508.08940","created_at":"2026-07-05T11:52:37.067874+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.08940v1","created_at":"2026-07-05T11:52:37.067874+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.08940","created_at":"2026-07-05T11:52:37.067874+00:00"},{"alias_kind":"pith_short_12","alias_value":"N6PCWXV7OGXO","created_at":"2026-07-05T11:52:37.067874+00:00"},{"alias_kind":"pith_short_16","alias_value":"N6PCWXV7OGXO4EIG","created_at":"2026-07-05T11:52:37.067874+00:00"},{"alias_kind":"pith_short_8","alias_value":"N6PCWXV7","created_at":"2026-07-05T11:52:37.067874+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2601.10348","citing_title":"Training-Trajectory-Aware Token Selection","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02913","citing_title":"Generate, Filter, Control, Replay: A Comprehensive Survey of Rollout Strategies for LLM Reinforcement Learning","ref_index":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/N6PCWXV7OGXO4EIGFTKTZJUEW7","json":"https://pith.science/pith/N6PCWXV7OGXO4EIGFTKTZJUEW7.json","graph_json":"https://pith.science/api/pith-number/N6PCWXV7OGXO4EIGFTKTZJUEW7/graph.json","events_json":"https://pith.science/api/pith-number/N6PCWXV7OGXO4EIGFTKTZJUEW7/events.json","paper":"https://pith.science/paper/N6PCWXV7"},"agent_actions":{"view_html":"https://pith.science/pith/N6PCWXV7OGXO4EIGFTKTZJUEW7","download_json":"https://pith.science/pith/N6PCWXV7OGXO4EIGFTKTZJUEW7.json","view_paper":"https://pith.science/paper/N6PCWXV7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.08940&json=true","fetch_graph":"https://pith.science/api/pith-number/N6PCWXV7OGXO4EIGFTKTZJUEW7/graph.json","fetch_events":"https://pith.science/api/pith-number/N6PCWXV7OGXO4EIGFTKTZJUEW7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/N6PCWXV7OGXO4EIGFTKTZJUEW7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/N6PCWXV7OGXO4EIGFTKTZJUEW7/action/storage_attestation","attest_author":"https://pith.science/pith/N6PCWXV7OGXO4EIGFTKTZJUEW7/action/author_attestation","sign_citation":"https://pith.science/pith/N6PCWXV7OGXO4EIGFTKTZJUEW7/action/citation_signature","submit_replication":"https://pith.science/pith/N6PCWXV7OGXO4EIGFTKTZJUEW7/action/replication_record"}},"created_at":"2026-07-05T11:52:37.067874+00:00","updated_at":"2026-07-05T11:52:37.067874+00:00"}