{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:VE4MLESY6PFXGGH7JVIUBUVH4B","short_pith_number":"pith:VE4MLESY","schema_version":"1.0","canonical_sha256":"a938c59258f3cb7318ff4d5140d2a7e041c3294e8e15f54ff0380e8c29275319","source":{"kind":"arxiv","id":"2501.19324","version":3},"attestation_state":"computed","paper":{"title":"Reward-Guided Speculative Decoding for Efficient LLM Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Baohao Liao, Caiming Xiong, Christof Monz, Doyen Sahoo, HanZe Dong, Junnan Li, Silvio Savarese, Yuhui Xu","submitted_at":"2025-01-31T17:19:57Z","abstract_excerpt":"We introduce Reward-Guided Speculative Decoding (RSD), a novel framework aimed at improving the efficiency of inference in large language models (LLMs). RSD synergistically combines a lightweight draft model with a more powerful target model, incorporating a controlled bias to prioritize high-reward outputs, in contrast to existing speculative decoding methods that enforce strict unbiasedness. RSD employs a process reward model to evaluate intermediate decoding steps and dynamically decide whether to invoke the target model, optimizing the trade-off between computational cost and output qualit"},"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":"2501.19324","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-01-31T17:19:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"db575acf30e99ae7390268d1bd0c381a380a5e5cbe88f810f2fd1e508663ebb7","abstract_canon_sha256":"34e90a4325f072689909425498170f533fd4aea627ca74ec505492f943b250a7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:27:23.764374Z","signature_b64":"CCJx/thvck4D4FaHzoYpZZRPXrhaXt8pzgmZWv+m3ESIipFEOjRWKsVbg+9VH9FYTSc2H2Jh1wn1fREFdHX+AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a938c59258f3cb7318ff4d5140d2a7e041c3294e8e15f54ff0380e8c29275319","last_reissued_at":"2026-07-05T11:27:23.763810Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:27:23.763810Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reward-Guided Speculative Decoding for Efficient LLM Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Baohao Liao, Caiming Xiong, Christof Monz, Doyen Sahoo, HanZe Dong, Junnan Li, Silvio Savarese, Yuhui Xu","submitted_at":"2025-01-31T17:19:57Z","abstract_excerpt":"We introduce Reward-Guided Speculative Decoding (RSD), a novel framework aimed at improving the efficiency of inference in large language models (LLMs). RSD synergistically combines a lightweight draft model with a more powerful target model, incorporating a controlled bias to prioritize high-reward outputs, in contrast to existing speculative decoding methods that enforce strict unbiasedness. RSD employs a process reward model to evaluate intermediate decoding steps and dynamically decide whether to invoke the target model, optimizing the trade-off between computational cost and output qualit"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.19324","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/2501.19324/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":"2501.19324","created_at":"2026-07-05T11:27:23.763875+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.19324v3","created_at":"2026-07-05T11:27:23.763875+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.19324","created_at":"2026-07-05T11:27:23.763875+00:00"},{"alias_kind":"pith_short_12","alias_value":"VE4MLESY6PFX","created_at":"2026-07-05T11:27:23.763875+00:00"},{"alias_kind":"pith_short_16","alias_value":"VE4MLESY6PFXGGH7","created_at":"2026-07-05T11:27:23.763875+00:00"},{"alias_kind":"pith_short_8","alias_value":"VE4MLESY","created_at":"2026-07-05T11:27:23.763875+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":14,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25354","citing_title":"Efficient and Trainable Language Model Test-Time Scaling via Local Branch Routing","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11470","citing_title":"The Periodic Table of LLM Reasoning: A Structured Survey of Reasoning Paradigms, Methods, and Failure Modes","ref_index":145,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03503","citing_title":"ThoughtFold: Folding Reasoning Chains via Introspective Preference Learning","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.25354","citing_title":"Efficient and Trainable Language Model Test-Time Scaling via Local Branch Routing","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30217","citing_title":"Before Thinking, Learn to Decide: Proactive Routing for Efficient Visual Reasoning","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10403","citing_title":"KCSAT-ML: Probing Reasoning Models with Nationwide-Cohort Human Difficulty","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14978","citing_title":"Performance-Driven Policy Optimization for Speculative Decoding with Adaptive Windowing","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2505.15436","citing_title":"Adaptive Chain-of-Focus Reasoning via Dynamic Visual Search and Zooming for Efficient VLMs","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10195","citing_title":"Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14098","citing_title":"Pause and Reflect: Conformal Aggregation for Chain-of-Thought Reasoning","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2503.16419","citing_title":"Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models","ref_index":100,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10195","citing_title":"Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration","ref_index":33,"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":115,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06116","citing_title":"Policy-Guided Stepwise Model Routing for Cost-Effective Reasoning","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VE4MLESY6PFXGGH7JVIUBUVH4B","json":"https://pith.science/pith/VE4MLESY6PFXGGH7JVIUBUVH4B.json","graph_json":"https://pith.science/api/pith-number/VE4MLESY6PFXGGH7JVIUBUVH4B/graph.json","events_json":"https://pith.science/api/pith-number/VE4MLESY6PFXGGH7JVIUBUVH4B/events.json","paper":"https://pith.science/paper/VE4MLESY"},"agent_actions":{"view_html":"https://pith.science/pith/VE4MLESY6PFXGGH7JVIUBUVH4B","download_json":"https://pith.science/pith/VE4MLESY6PFXGGH7JVIUBUVH4B.json","view_paper":"https://pith.science/paper/VE4MLESY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.19324&json=true","fetch_graph":"https://pith.science/api/pith-number/VE4MLESY6PFXGGH7JVIUBUVH4B/graph.json","fetch_events":"https://pith.science/api/pith-number/VE4MLESY6PFXGGH7JVIUBUVH4B/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VE4MLESY6PFXGGH7JVIUBUVH4B/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VE4MLESY6PFXGGH7JVIUBUVH4B/action/storage_attestation","attest_author":"https://pith.science/pith/VE4MLESY6PFXGGH7JVIUBUVH4B/action/author_attestation","sign_citation":"https://pith.science/pith/VE4MLESY6PFXGGH7JVIUBUVH4B/action/citation_signature","submit_replication":"https://pith.science/pith/VE4MLESY6PFXGGH7JVIUBUVH4B/action/replication_record"}},"created_at":"2026-07-05T11:27:23.763875+00:00","updated_at":"2026-07-05T11:27:23.763875+00:00"}