{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5JP5U7OCJN7VNV54JA5WRCGFP3","short_pith_number":"pith:5JP5U7OC","schema_version":"1.0","canonical_sha256":"ea5fda7dc24b7f56d7bc483b6888c57efe80f51c27e35e21be781a67393de3d5","source":{"kind":"arxiv","id":"2505.07961","version":3},"attestation_state":"computed","paper":{"title":"Making Small Language Models Efficient Reasoners: Intervention, Supervision, Reinforcement","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chenshun Ni, Jiasi Chen, Samet Oymak, Xuechen Zhang, Zijian Huang, Ziyang Xiong","submitted_at":"2025-05-12T18:04:39Z","abstract_excerpt":"Recent research enhances language model reasoning by scaling test-time compute via longer chain-of-thought traces. This often improves accuracy but also introduces redundancy and high computational cost, especially for small language models distilled with supervised fine-tuning (SFT). In this work, we propose new algorithms to improve token-efficient reasoning with small-scale models by effectively trading off accuracy and computation. We first show that the post-SFT model fails to determine the optimal stopping point of the reasoning process, resulting in verbose and repetitive outputs. Verbo"},"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":"2505.07961","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-12T18:04:39Z","cross_cats_sorted":[],"title_canon_sha256":"ac10a0a9fdac1bbbbd52ee64bf3e7d6e4ea9293822183661041897bd8a8b90a0","abstract_canon_sha256":"6b9946174c400e823aa6f32cf3aa38df5fbf697356d241e0405cdf88d577480d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:08:20.341515Z","signature_b64":"Ls1jalC638F61N1V/93aDlxSVeMMT4UTZ7XF5I5TY6iodvpr+n9jysKQmrS7U5QkSquEagi1nYg8g5CnndVcAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ea5fda7dc24b7f56d7bc483b6888c57efe80f51c27e35e21be781a67393de3d5","last_reissued_at":"2026-07-05T11:08:20.341015Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:08:20.341015Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Making Small Language Models Efficient Reasoners: Intervention, Supervision, Reinforcement","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chenshun Ni, Jiasi Chen, Samet Oymak, Xuechen Zhang, Zijian Huang, Ziyang Xiong","submitted_at":"2025-05-12T18:04:39Z","abstract_excerpt":"Recent research enhances language model reasoning by scaling test-time compute via longer chain-of-thought traces. This often improves accuracy but also introduces redundancy and high computational cost, especially for small language models distilled with supervised fine-tuning (SFT). In this work, we propose new algorithms to improve token-efficient reasoning with small-scale models by effectively trading off accuracy and computation. We first show that the post-SFT model fails to determine the optimal stopping point of the reasoning process, resulting in verbose and repetitive outputs. Verbo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.07961","kind":"arxiv","version":3},"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/2505.07961/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":"2505.07961","created_at":"2026-07-05T11:08:20.341074+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.07961v3","created_at":"2026-07-05T11:08:20.341074+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.07961","created_at":"2026-07-05T11:08:20.341074+00:00"},{"alias_kind":"pith_short_12","alias_value":"5JP5U7OCJN7V","created_at":"2026-07-05T11:08:20.341074+00:00"},{"alias_kind":"pith_short_16","alias_value":"5JP5U7OCJN7VNV54","created_at":"2026-07-05T11:08:20.341074+00:00"},{"alias_kind":"pith_short_8","alias_value":"5JP5U7OC","created_at":"2026-07-05T11:08:20.341074+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.15604","citing_title":"VSPO: Vector-Steered Policy Optimization for Behavioral Control","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2506.17788","citing_title":"Bayesian Social Deduction with Graph-Informed Language Models","ref_index":76,"is_internal_anchor":false},{"citing_arxiv_id":"2503.16419","citing_title":"Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models","ref_index":240,"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":233,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5JP5U7OCJN7VNV54JA5WRCGFP3","json":"https://pith.science/pith/5JP5U7OCJN7VNV54JA5WRCGFP3.json","graph_json":"https://pith.science/api/pith-number/5JP5U7OCJN7VNV54JA5WRCGFP3/graph.json","events_json":"https://pith.science/api/pith-number/5JP5U7OCJN7VNV54JA5WRCGFP3/events.json","paper":"https://pith.science/paper/5JP5U7OC"},"agent_actions":{"view_html":"https://pith.science/pith/5JP5U7OCJN7VNV54JA5WRCGFP3","download_json":"https://pith.science/pith/5JP5U7OCJN7VNV54JA5WRCGFP3.json","view_paper":"https://pith.science/paper/5JP5U7OC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.07961&json=true","fetch_graph":"https://pith.science/api/pith-number/5JP5U7OCJN7VNV54JA5WRCGFP3/graph.json","fetch_events":"https://pith.science/api/pith-number/5JP5U7OCJN7VNV54JA5WRCGFP3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5JP5U7OCJN7VNV54JA5WRCGFP3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5JP5U7OCJN7VNV54JA5WRCGFP3/action/storage_attestation","attest_author":"https://pith.science/pith/5JP5U7OCJN7VNV54JA5WRCGFP3/action/author_attestation","sign_citation":"https://pith.science/pith/5JP5U7OCJN7VNV54JA5WRCGFP3/action/citation_signature","submit_replication":"https://pith.science/pith/5JP5U7OCJN7VNV54JA5WRCGFP3/action/replication_record"}},"created_at":"2026-07-05T11:08:20.341074+00:00","updated_at":"2026-07-05T11:08:20.341074+00:00"}