{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:C2YUIR3QCVYINEVQM77YI5ZQW7","short_pith_number":"pith:C2YUIR3Q","schema_version":"1.0","canonical_sha256":"16b144477015708692b067ff847730b7c221fab473c6c769707125d5be42bf02","source":{"kind":"arxiv","id":"2504.12329","version":2},"attestation_state":"computed","paper":{"title":"Speculative Thinking: Enhancing Small-Model Reasoning with Large Model Guidance at Inference Time","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Vipin Chaudhary, Wang Yang, Xiang Yue, Xiaotian Han","submitted_at":"2025-04-12T21:25:32Z","abstract_excerpt":"Recent advances leverage post-training to enhance model reasoning performance, which typically requires costly training pipelines and still suffers from inefficient, overly lengthy outputs. We introduce Speculative Thinking, a training-free framework that enables large reasoning models to guide smaller ones during inference at the reasoning level, distinct from speculative decoding, which operates at the token level. Our approach is based on two observations: (1) reasoning-supportive tokens such as \"wait\" frequently appear after structural delimiters like \"\\n\\n\", serving as signals for reflect"},"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":"2504.12329","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-12T21:25:32Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ec3390046a441c16db5791706b46291dd8c18a1e8d022416f66efad14153bdcc","abstract_canon_sha256":"901b8423bb9fdc4fa879148e7c342032626909467134f732e2d634addaeab20a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-04T01:08:28.730728Z","signature_b64":"y7Kev0rXDs+y678oWbJGUbLCiojs5ZmL4IqU6oQJoLQPonDB5/sZ+r6fX2KDqGCu3bl2OPD0lVOv/ZeIlt2FAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"16b144477015708692b067ff847730b7c221fab473c6c769707125d5be42bf02","last_reissued_at":"2026-06-04T01:08:28.730158Z","signature_status":"signed_v1","first_computed_at":"2026-06-04T01:08:28.730158Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Speculative Thinking: Enhancing Small-Model Reasoning with Large Model Guidance at Inference Time","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Vipin Chaudhary, Wang Yang, Xiang Yue, Xiaotian Han","submitted_at":"2025-04-12T21:25:32Z","abstract_excerpt":"Recent advances leverage post-training to enhance model reasoning performance, which typically requires costly training pipelines and still suffers from inefficient, overly lengthy outputs. We introduce Speculative Thinking, a training-free framework that enables large reasoning models to guide smaller ones during inference at the reasoning level, distinct from speculative decoding, which operates at the token level. Our approach is based on two observations: (1) reasoning-supportive tokens such as \"wait\" frequently appear after structural delimiters like \"\\n\\n\", serving as signals for reflect"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.12329","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/2504.12329/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":"2504.12329","created_at":"2026-06-04T01:08:28.730232+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.12329v2","created_at":"2026-06-04T01:08:28.730232+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.12329","created_at":"2026-06-04T01:08:28.730232+00:00"},{"alias_kind":"pith_short_12","alias_value":"C2YUIR3QCVYI","created_at":"2026-06-04T01:08:28.730232+00:00"},{"alias_kind":"pith_short_16","alias_value":"C2YUIR3QCVYINEVQ","created_at":"2026-06-04T01:08:28.730232+00:00"},{"alias_kind":"pith_short_8","alias_value":"C2YUIR3Q","created_at":"2026-06-04T01:08:28.730232+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":5,"sample":[{"citing_arxiv_id":"2605.10195","citing_title":"Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration","ref_index":63,"is_internal_anchor":true},{"citing_arxiv_id":"2503.16419","citing_title":"Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models","ref_index":213,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10195","citing_title":"Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration","ref_index":63,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06165","citing_title":"Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost","ref_index":195,"is_internal_anchor":true},{"citing_arxiv_id":"2604.18839","citing_title":"One Step Forward and K Steps Back: Better Reasoning with Denoising Recursion Models","ref_index":140,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/C2YUIR3QCVYINEVQM77YI5ZQW7","json":"https://pith.science/pith/C2YUIR3QCVYINEVQM77YI5ZQW7.json","graph_json":"https://pith.science/api/pith-number/C2YUIR3QCVYINEVQM77YI5ZQW7/graph.json","events_json":"https://pith.science/api/pith-number/C2YUIR3QCVYINEVQM77YI5ZQW7/events.json","paper":"https://pith.science/paper/C2YUIR3Q"},"agent_actions":{"view_html":"https://pith.science/pith/C2YUIR3QCVYINEVQM77YI5ZQW7","download_json":"https://pith.science/pith/C2YUIR3QCVYINEVQM77YI5ZQW7.json","view_paper":"https://pith.science/paper/C2YUIR3Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.12329&json=true","fetch_graph":"https://pith.science/api/pith-number/C2YUIR3QCVYINEVQM77YI5ZQW7/graph.json","fetch_events":"https://pith.science/api/pith-number/C2YUIR3QCVYINEVQM77YI5ZQW7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/C2YUIR3QCVYINEVQM77YI5ZQW7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/C2YUIR3QCVYINEVQM77YI5ZQW7/action/storage_attestation","attest_author":"https://pith.science/pith/C2YUIR3QCVYINEVQM77YI5ZQW7/action/author_attestation","sign_citation":"https://pith.science/pith/C2YUIR3QCVYINEVQM77YI5ZQW7/action/citation_signature","submit_replication":"https://pith.science/pith/C2YUIR3QCVYINEVQM77YI5ZQW7/action/replication_record"}},"created_at":"2026-06-04T01:08:28.730232+00:00","updated_at":"2026-06-04T01:08:28.730232+00:00"}