{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:J6VMPWAJETCMQYRCICP6MJNROL","short_pith_number":"pith:J6VMPWAJ","schema_version":"1.0","canonical_sha256":"4faac7d80924c4c86222409fe625b172d20d5a8f217b36fc2aa78e4380236d1a","source":{"kind":"arxiv","id":"2309.08168","version":2},"attestation_state":"computed","paper":{"title":"Draft & Verify: Lossless Large Language Model Acceleration via Self-Speculative Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Gang Chen, Huan Li, Jue Wang, Jun Zhang, Ke Chen, Lidan Shou, Sharad Mehrotra","submitted_at":"2023-09-15T05:34:32Z","abstract_excerpt":"We present a novel inference scheme, self-speculative decoding, for accelerating Large Language Models (LLMs) without the need for an auxiliary model. This approach is characterized by a two-stage process: drafting and verification. The drafting stage generates draft tokens at a slightly lower quality but more quickly, which is achieved by selectively skipping certain intermediate layers during drafting. Subsequently, the verification stage employs the original LLM to validate those draft output tokens in one forward pass. This process ensures the final output remains identical to that produce"},"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":"2309.08168","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-15T05:34:32Z","cross_cats_sorted":[],"title_canon_sha256":"76f9bbbb77b5e407a82cb8114d7bca23881e396f8ab70f48b09e534c28bc12e3","abstract_canon_sha256":"be18ba807cc8a968536b079faae2c4fbaf2bb7d221684860e0ce769488ade337"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:11:13.810787Z","signature_b64":"cYw8jWXyp7c1rENVo6YkWuWaxwv+cekkatYrY8zgiIEQNe0+jvksAgxWbQVQVP9q40q+GxcfW7Gj5hbgEOpLCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4faac7d80924c4c86222409fe625b172d20d5a8f217b36fc2aa78e4380236d1a","last_reissued_at":"2026-07-05T10:11:13.810354Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:11:13.810354Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Draft & Verify: Lossless Large Language Model Acceleration via Self-Speculative Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Gang Chen, Huan Li, Jue Wang, Jun Zhang, Ke Chen, Lidan Shou, Sharad Mehrotra","submitted_at":"2023-09-15T05:34:32Z","abstract_excerpt":"We present a novel inference scheme, self-speculative decoding, for accelerating Large Language Models (LLMs) without the need for an auxiliary model. This approach is characterized by a two-stage process: drafting and verification. The drafting stage generates draft tokens at a slightly lower quality but more quickly, which is achieved by selectively skipping certain intermediate layers during drafting. Subsequently, the verification stage employs the original LLM to validate those draft output tokens in one forward pass. This process ensures the final output remains identical to that produce"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.08168","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/2309.08168/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":"2309.08168","created_at":"2026-07-05T10:11:13.810425+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.08168v2","created_at":"2026-07-05T10:11:13.810425+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.08168","created_at":"2026-07-05T10:11:13.810425+00:00"},{"alias_kind":"pith_short_12","alias_value":"J6VMPWAJETCM","created_at":"2026-07-05T10:11:13.810425+00:00"},{"alias_kind":"pith_short_16","alias_value":"J6VMPWAJETCMQYRC","created_at":"2026-07-05T10:11:13.810425+00:00"},{"alias_kind":"pith_short_8","alias_value":"J6VMPWAJ","created_at":"2026-07-05T10:11:13.810425+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.28678","citing_title":"DREAM-R: Multimodal Speculative Reasoning with RL-Based Refined Drafting, Precise Verification, and Fully Parallel Execution","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00535","citing_title":"DREAM-S: Speculative Decoding with Searchable Drafting and Target-Aware Refinement for Multimodal Generation","ref_index":116,"is_internal_anchor":false},{"citing_arxiv_id":"2603.01581","citing_title":"KERV: Kinematic-Rectified Speculative Decoding for Embodied VLA Models","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2404.14294","citing_title":"A Survey on Efficient Inference for Large Language Models","ref_index":239,"is_internal_anchor":false},{"citing_arxiv_id":"2401.15077","citing_title":"EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty","ref_index":83,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/J6VMPWAJETCMQYRCICP6MJNROL","json":"https://pith.science/pith/J6VMPWAJETCMQYRCICP6MJNROL.json","graph_json":"https://pith.science/api/pith-number/J6VMPWAJETCMQYRCICP6MJNROL/graph.json","events_json":"https://pith.science/api/pith-number/J6VMPWAJETCMQYRCICP6MJNROL/events.json","paper":"https://pith.science/paper/J6VMPWAJ"},"agent_actions":{"view_html":"https://pith.science/pith/J6VMPWAJETCMQYRCICP6MJNROL","download_json":"https://pith.science/pith/J6VMPWAJETCMQYRCICP6MJNROL.json","view_paper":"https://pith.science/paper/J6VMPWAJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.08168&json=true","fetch_graph":"https://pith.science/api/pith-number/J6VMPWAJETCMQYRCICP6MJNROL/graph.json","fetch_events":"https://pith.science/api/pith-number/J6VMPWAJETCMQYRCICP6MJNROL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/J6VMPWAJETCMQYRCICP6MJNROL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/J6VMPWAJETCMQYRCICP6MJNROL/action/storage_attestation","attest_author":"https://pith.science/pith/J6VMPWAJETCMQYRCICP6MJNROL/action/author_attestation","sign_citation":"https://pith.science/pith/J6VMPWAJETCMQYRCICP6MJNROL/action/citation_signature","submit_replication":"https://pith.science/pith/J6VMPWAJETCMQYRCICP6MJNROL/action/replication_record"}},"created_at":"2026-07-05T10:11:13.810425+00:00","updated_at":"2026-07-05T10:11:13.810425+00:00"}