{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2EAINMEFBZZZXYOVH2HBTJBZVQ","short_pith_number":"pith:2EAINMEF","schema_version":"1.0","canonical_sha256":"d10086b0850e739be1d53e8e19a439ac1eabfb755650ff4bcfc88c0c2ba716f6","source":{"kind":"arxiv","id":"2507.12759","version":1},"attestation_state":"computed","paper":{"title":"Logit Arithmetic Elicits Long Reasoning Capabilities Without Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ayoung Lee, Farima Fatahi Bayat, Lechen Zhang, Lu Wang, Muhammad Khalifa, Xinliang Frederick Zhang, Xin Liu, Yunxiang Zhang","submitted_at":"2025-07-17T03:31:36Z","abstract_excerpt":"Large reasoning models (LRMs) can do complex reasoning via long chain-of-thought (CoT) involving cognitive strategies such as backtracking and self-correction. Recent studies suggest that some models inherently possess these long reasoning abilities, which may be unlocked via extra training. Our work first investigates whether we can elicit such behavior without any training. To this end, we propose a decoding-time approach, ThinkLogit, which utilizes logits arithmetic (Liu et al., 2024) to tune a target large LM for long reasoning using a substantially smaller model as guider. We then show th"},"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":"2507.12759","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-17T03:31:36Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dc4fca17ac4cc83f59e558c9bc6025e99b487845f27768ffeeeb96ba45a5acf9","abstract_canon_sha256":"0330e586f3dfb8afaccaf1f16839912281c96630bae90800ad86bafe7455188c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:38:38.595758Z","signature_b64":"0sr8u4u75lTnEF1zumNkdwwUYfVh0vDDKH2KIZNtgw89DjDSl6nNSWIOJQ87ZI3ZnqoXqDEvcPNVQPx6rBvtDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d10086b0850e739be1d53e8e19a439ac1eabfb755650ff4bcfc88c0c2ba716f6","last_reissued_at":"2026-07-05T11:38:38.595251Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:38:38.595251Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Logit Arithmetic Elicits Long Reasoning Capabilities Without Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ayoung Lee, Farima Fatahi Bayat, Lechen Zhang, Lu Wang, Muhammad Khalifa, Xinliang Frederick Zhang, Xin Liu, Yunxiang Zhang","submitted_at":"2025-07-17T03:31:36Z","abstract_excerpt":"Large reasoning models (LRMs) can do complex reasoning via long chain-of-thought (CoT) involving cognitive strategies such as backtracking and self-correction. Recent studies suggest that some models inherently possess these long reasoning abilities, which may be unlocked via extra training. Our work first investigates whether we can elicit such behavior without any training. To this end, we propose a decoding-time approach, ThinkLogit, which utilizes logits arithmetic (Liu et al., 2024) to tune a target large LM for long reasoning using a substantially smaller model as guider. We then show th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.12759","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/2507.12759/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":"2507.12759","created_at":"2026-07-05T11:38:38.595311+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.12759v1","created_at":"2026-07-05T11:38:38.595311+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.12759","created_at":"2026-07-05T11:38:38.595311+00:00"},{"alias_kind":"pith_short_12","alias_value":"2EAINMEFBZZZ","created_at":"2026-07-05T11:38:38.595311+00:00"},{"alias_kind":"pith_short_16","alias_value":"2EAINMEFBZZZXYOV","created_at":"2026-07-05T11:38:38.595311+00:00"},{"alias_kind":"pith_short_8","alias_value":"2EAINMEF","created_at":"2026-07-05T11:38:38.595311+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2EAINMEFBZZZXYOVH2HBTJBZVQ","json":"https://pith.science/pith/2EAINMEFBZZZXYOVH2HBTJBZVQ.json","graph_json":"https://pith.science/api/pith-number/2EAINMEFBZZZXYOVH2HBTJBZVQ/graph.json","events_json":"https://pith.science/api/pith-number/2EAINMEFBZZZXYOVH2HBTJBZVQ/events.json","paper":"https://pith.science/paper/2EAINMEF"},"agent_actions":{"view_html":"https://pith.science/pith/2EAINMEFBZZZXYOVH2HBTJBZVQ","download_json":"https://pith.science/pith/2EAINMEFBZZZXYOVH2HBTJBZVQ.json","view_paper":"https://pith.science/paper/2EAINMEF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.12759&json=true","fetch_graph":"https://pith.science/api/pith-number/2EAINMEFBZZZXYOVH2HBTJBZVQ/graph.json","fetch_events":"https://pith.science/api/pith-number/2EAINMEFBZZZXYOVH2HBTJBZVQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2EAINMEFBZZZXYOVH2HBTJBZVQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2EAINMEFBZZZXYOVH2HBTJBZVQ/action/storage_attestation","attest_author":"https://pith.science/pith/2EAINMEFBZZZXYOVH2HBTJBZVQ/action/author_attestation","sign_citation":"https://pith.science/pith/2EAINMEFBZZZXYOVH2HBTJBZVQ/action/citation_signature","submit_replication":"https://pith.science/pith/2EAINMEFBZZZXYOVH2HBTJBZVQ/action/replication_record"}},"created_at":"2026-07-05T11:38:38.595311+00:00","updated_at":"2026-07-05T11:38:38.595311+00:00"}