{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:6H7OZ3XCU7RRK3BQG4C26RO773","short_pith_number":"pith:6H7OZ3XC","schema_version":"1.0","canonical_sha256":"f1feeceee2a7e3156c303705af45dffee438527e1372c7dffdf59c2f73b70213","source":{"kind":"arxiv","id":"2506.10822","version":1},"attestation_state":"computed","paper":{"title":"ReCUT: Balancing Reasoning Length and Accuracy in LLMs via Stepwise Trails and Preference Optimization","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chunyi Peng, Furong Peng, Ge Yu, Qi Shi, Shuo Wang, Xinze Li, Yifan Ji, Yukun Yan, Zhenghao Liu, Zhensheng Jin","submitted_at":"2025-06-12T15:43:01Z","abstract_excerpt":"Recent advances in Chain-of-Thought (CoT) prompting have substantially improved the reasoning capabilities of Large Language Models (LLMs). However, these methods often suffer from overthinking, leading to unnecessarily lengthy or redundant reasoning traces. Existing approaches attempt to mitigate this issue through curating multiple reasoning chains for training LLMs, but their effectiveness is often constrained by the quality of the generated data and prone to overfitting. To address the challenge, we propose Reasoning Compression ThroUgh Stepwise Trials (ReCUT), a novel method aimed at bala"},"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":"2506.10822","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-12T15:43:01Z","cross_cats_sorted":[],"title_canon_sha256":"2424b58344ae69e8c667a2ab028def3649c1dc8507fcd717dae36c665cc8882a","abstract_canon_sha256":"4fd0c78c3603d0d65e127aa71ef5f8978921cfd9806a96322cf384f937f1c0e9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:20:31.202413Z","signature_b64":"ZFDh+hFaMC9H9YsIbHTLdLKswebA0C7uuLAEGebb/SERrXF7ECqHORXM7coZsQ67hRyYj3XEos2xS7W6B27JBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f1feeceee2a7e3156c303705af45dffee438527e1372c7dffdf59c2f73b70213","last_reissued_at":"2026-07-05T11:20:31.201900Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:20:31.201900Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ReCUT: Balancing Reasoning Length and Accuracy in LLMs via Stepwise Trails and Preference Optimization","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chunyi Peng, Furong Peng, Ge Yu, Qi Shi, Shuo Wang, Xinze Li, Yifan Ji, Yukun Yan, Zhenghao Liu, Zhensheng Jin","submitted_at":"2025-06-12T15:43:01Z","abstract_excerpt":"Recent advances in Chain-of-Thought (CoT) prompting have substantially improved the reasoning capabilities of Large Language Models (LLMs). However, these methods often suffer from overthinking, leading to unnecessarily lengthy or redundant reasoning traces. Existing approaches attempt to mitigate this issue through curating multiple reasoning chains for training LLMs, but their effectiveness is often constrained by the quality of the generated data and prone to overfitting. To address the challenge, we propose Reasoning Compression ThroUgh Stepwise Trials (ReCUT), a novel method aimed at bala"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.10822","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/2506.10822/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":"2506.10822","created_at":"2026-07-05T11:20:31.201960+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.10822v1","created_at":"2026-07-05T11:20:31.201960+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.10822","created_at":"2026-07-05T11:20:31.201960+00:00"},{"alias_kind":"pith_short_12","alias_value":"6H7OZ3XCU7RR","created_at":"2026-07-05T11:20:31.201960+00:00"},{"alias_kind":"pith_short_16","alias_value":"6H7OZ3XCU7RRK3BQ","created_at":"2026-07-05T11:20:31.201960+00:00"},{"alias_kind":"pith_short_8","alias_value":"6H7OZ3XC","created_at":"2026-07-05T11:20:31.201960+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22716","citing_title":"Beyond Penalizing Mistakes: Stabilizing Efficiency Training in Large Reasoning Models via Adaptive Correct-Only Rewards","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6H7OZ3XCU7RRK3BQG4C26RO773","json":"https://pith.science/pith/6H7OZ3XCU7RRK3BQG4C26RO773.json","graph_json":"https://pith.science/api/pith-number/6H7OZ3XCU7RRK3BQG4C26RO773/graph.json","events_json":"https://pith.science/api/pith-number/6H7OZ3XCU7RRK3BQG4C26RO773/events.json","paper":"https://pith.science/paper/6H7OZ3XC"},"agent_actions":{"view_html":"https://pith.science/pith/6H7OZ3XCU7RRK3BQG4C26RO773","download_json":"https://pith.science/pith/6H7OZ3XCU7RRK3BQG4C26RO773.json","view_paper":"https://pith.science/paper/6H7OZ3XC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.10822&json=true","fetch_graph":"https://pith.science/api/pith-number/6H7OZ3XCU7RRK3BQG4C26RO773/graph.json","fetch_events":"https://pith.science/api/pith-number/6H7OZ3XCU7RRK3BQG4C26RO773/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6H7OZ3XCU7RRK3BQG4C26RO773/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6H7OZ3XCU7RRK3BQG4C26RO773/action/storage_attestation","attest_author":"https://pith.science/pith/6H7OZ3XCU7RRK3BQG4C26RO773/action/author_attestation","sign_citation":"https://pith.science/pith/6H7OZ3XCU7RRK3BQG4C26RO773/action/citation_signature","submit_replication":"https://pith.science/pith/6H7OZ3XCU7RRK3BQG4C26RO773/action/replication_record"}},"created_at":"2026-07-05T11:20:31.201960+00:00","updated_at":"2026-07-05T11:20:31.201960+00:00"}