{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5BCTBO23BCBSLOYBDNIYIJVOZJ","short_pith_number":"pith:5BCTBO23","schema_version":"1.0","canonical_sha256":"e84530bb5b088325bb011b518426aeca6464845d09c9e2576f15a1652e77011e","source":{"kind":"arxiv","id":"2510.07962","version":2},"attestation_state":"computed","paper":{"title":"LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chao Huang, Jingyuan Wang, Yankai Chen, Zhonghang Li","submitted_at":"2025-10-09T08:55:12Z","abstract_excerpt":"Large language models (LLMs) have demonstrated remarkable progress in reasoning, often through supervised fine-tuning (SFT). However, SFT is resource-intensive, relying on large curated datasets, rejection-sampled demonstrations, and uniform optimization across all tokens, even though only a fraction carry meaningful learning value. In this work, we explore a counterintuitive idea: can smaller language models (SLMs) teach larger language models (LLMs) by revealing high-value reasoning moments that reflect the latter's unique strength? We propose LightReasoner, a novel framework that leverages "},"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":"2510.07962","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-10-09T08:55:12Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"0265e109227a71c129e13cc5ec8aa6c0c391b1c6a81e99414949a363440e81d8","abstract_canon_sha256":"8ef0af77c9bdedc488c34c4c498cc24bebc075d765df843fc714db10f677cfcc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:03:47.964097Z","signature_b64":"N7TtGZaMxi4rA2CA4vF1umLCys5GfAx6i6aOmwSUs8XHej1AzbSnwGRUJjIldlx9u910X2EI/puLi3i69vPYCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e84530bb5b088325bb011b518426aeca6464845d09c9e2576f15a1652e77011e","last_reissued_at":"2026-05-22T01:03:47.962960Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:03:47.962960Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chao Huang, Jingyuan Wang, Yankai Chen, Zhonghang Li","submitted_at":"2025-10-09T08:55:12Z","abstract_excerpt":"Large language models (LLMs) have demonstrated remarkable progress in reasoning, often through supervised fine-tuning (SFT). However, SFT is resource-intensive, relying on large curated datasets, rejection-sampled demonstrations, and uniform optimization across all tokens, even though only a fraction carry meaningful learning value. In this work, we explore a counterintuitive idea: can smaller language models (SLMs) teach larger language models (LLMs) by revealing high-value reasoning moments that reflect the latter's unique strength? We propose LightReasoner, a novel framework that leverages "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.07962","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2510.07962/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":"2510.07962","created_at":"2026-05-22T01:03:47.963094+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.07962v2","created_at":"2026-05-22T01:03:47.963094+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.07962","created_at":"2026-05-22T01:03:47.963094+00:00"},{"alias_kind":"pith_short_12","alias_value":"5BCTBO23BCBS","created_at":"2026-05-22T01:03:47.963094+00:00"},{"alias_kind":"pith_short_16","alias_value":"5BCTBO23BCBSLOYB","created_at":"2026-05-22T01:03:47.963094+00:00"},{"alias_kind":"pith_short_8","alias_value":"5BCTBO23","created_at":"2026-05-22T01:03:47.963094+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.06377","citing_title":"The Master Key Hypothesis: Unlocking Cross-Model Capability Transfer via Linear Subspace Alignment","ref_index":74,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5BCTBO23BCBSLOYBDNIYIJVOZJ","json":"https://pith.science/pith/5BCTBO23BCBSLOYBDNIYIJVOZJ.json","graph_json":"https://pith.science/api/pith-number/5BCTBO23BCBSLOYBDNIYIJVOZJ/graph.json","events_json":"https://pith.science/api/pith-number/5BCTBO23BCBSLOYBDNIYIJVOZJ/events.json","paper":"https://pith.science/paper/5BCTBO23"},"agent_actions":{"view_html":"https://pith.science/pith/5BCTBO23BCBSLOYBDNIYIJVOZJ","download_json":"https://pith.science/pith/5BCTBO23BCBSLOYBDNIYIJVOZJ.json","view_paper":"https://pith.science/paper/5BCTBO23","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.07962&json=true","fetch_graph":"https://pith.science/api/pith-number/5BCTBO23BCBSLOYBDNIYIJVOZJ/graph.json","fetch_events":"https://pith.science/api/pith-number/5BCTBO23BCBSLOYBDNIYIJVOZJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5BCTBO23BCBSLOYBDNIYIJVOZJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5BCTBO23BCBSLOYBDNIYIJVOZJ/action/storage_attestation","attest_author":"https://pith.science/pith/5BCTBO23BCBSLOYBDNIYIJVOZJ/action/author_attestation","sign_citation":"https://pith.science/pith/5BCTBO23BCBSLOYBDNIYIJVOZJ/action/citation_signature","submit_replication":"https://pith.science/pith/5BCTBO23BCBSLOYBDNIYIJVOZJ/action/replication_record"}},"created_at":"2026-05-22T01:03:47.963094+00:00","updated_at":"2026-05-22T01:03:47.963094+00:00"}