{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:FIOEL4S7APCV3BFOPD3TDI2HIY","short_pith_number":"pith:FIOEL4S7","schema_version":"1.0","canonical_sha256":"2a1c45f25f03c55d84ae78f731a34746006161c12664e3636670c919570f97a0","source":{"kind":"arxiv","id":"2305.09781","version":4},"attestation_state":"computed","paper":{"title":"SpecInfer: Accelerating Generative Large Language Model Serving with Tree-based Speculative Inference and Verification","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DC","cs.LG"],"primary_cat":"cs.CL","authors_text":"Alan Zhu, Chunan Shi, Daiyaan Arfeen, Gabriele Oliaro, Lijie Yang, Rae Ying Yee Wong, Reyna Abhyankar, Xiaoxiang Shi, Xinhao Cheng, Xupeng Miao, Zeyu Wang, Zhengxin Zhang, Zhihao Jia, Zhihao Zhang, Zhuoming Chen","submitted_at":"2023-05-16T20:12:59Z","abstract_excerpt":"This paper introduces SpecInfer, a system that accelerates generative large language model (LLM) serving with tree-based speculative inference and verification. The key idea behind SpecInfer is leveraging small speculative models to predict the LLM's outputs; the predictions are organized as a token tree, whose nodes each represent a candidate token sequence. The correctness of all candidate token sequences represented by a token tree is verified against the LLM in parallel using a novel tree-based parallel decoding mechanism. SpecInfer uses an LLM as a token tree verifier instead of an increm"},"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":"2305.09781","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-16T20:12:59Z","cross_cats_sorted":["cs.DC","cs.LG"],"title_canon_sha256":"ab13a9de685889f20cf8f3ecbb9ac9d69991cef06cc9350541c3c5c1f83cbdd2","abstract_canon_sha256":"7a1ef1ac8c3fc6eeed746b42c6ebd95e35bb7ba2bb97c2923f948052ec3a3c36"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:02:43.547984Z","signature_b64":"NYEJpy8utzd+ZNsart64VFwI7zTJaLam0GrpEXCqi0LX5N6ScNKKqqbkHIdWWmUGNgUo0jIb1IuQI/cguUjLBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2a1c45f25f03c55d84ae78f731a34746006161c12664e3636670c919570f97a0","last_reissued_at":"2026-07-05T08:02:43.547514Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:02:43.547514Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SpecInfer: Accelerating Generative Large Language Model Serving with Tree-based Speculative Inference and Verification","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DC","cs.LG"],"primary_cat":"cs.CL","authors_text":"Alan Zhu, Chunan Shi, Daiyaan Arfeen, Gabriele Oliaro, Lijie Yang, Rae Ying Yee Wong, Reyna Abhyankar, Xiaoxiang Shi, Xinhao Cheng, Xupeng Miao, Zeyu Wang, Zhengxin Zhang, Zhihao Jia, Zhihao Zhang, Zhuoming Chen","submitted_at":"2023-05-16T20:12:59Z","abstract_excerpt":"This paper introduces SpecInfer, a system that accelerates generative large language model (LLM) serving with tree-based speculative inference and verification. The key idea behind SpecInfer is leveraging small speculative models to predict the LLM's outputs; the predictions are organized as a token tree, whose nodes each represent a candidate token sequence. The correctness of all candidate token sequences represented by a token tree is verified against the LLM in parallel using a novel tree-based parallel decoding mechanism. SpecInfer uses an LLM as a token tree verifier instead of an increm"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.09781","kind":"arxiv","version":4},"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/2305.09781/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":"2305.09781","created_at":"2026-07-05T08:02:43.547575+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.09781v4","created_at":"2026-07-05T08:02:43.547575+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.09781","created_at":"2026-07-05T08:02:43.547575+00:00"},{"alias_kind":"pith_short_12","alias_value":"FIOEL4S7APCV","created_at":"2026-07-05T08:02:43.547575+00:00"},{"alias_kind":"pith_short_16","alias_value":"FIOEL4S7APCV3BFO","created_at":"2026-07-05T08:02:43.547575+00:00"},{"alias_kind":"pith_short_8","alias_value":"FIOEL4S7","created_at":"2026-07-05T08:02:43.547575+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":19,"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":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.25097","citing_title":"Speculative Decoding at Temperature Zero: A Scoped Safety-Invariance Screen with a 48,072-Sample Expansion","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20295","citing_title":"Token-Operations-Oriented Inference Optimization Techniques for Large Models","ref_index":136,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12303","citing_title":"From 2D Grids to 1D Tokens: Reforming Shared Representations for Multimodal Image Fusion","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02091","citing_title":"DFlare: Scaling Up Draft Capacity for Block Diffusion Speculative Decoding","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26940","citing_title":"Select to Think: Unlocking SLM Potential with Local Sufficiency","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":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29707","citing_title":"Domino: Decoupling Causal Modeling from Autoregressive Drafting in Speculative Decoding","ref_index":21,"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":93,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15609","citing_title":"PSD: Pushing the Pareto Frontier of Diffusion LLMs via Parallel Speculative Decoding","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2602.06932","citing_title":"When RL Meets Adaptive Speculative Training: A Unified Training-Serving System","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2404.14294","citing_title":"A Survey on Efficient Inference for Large Language Models","ref_index":243,"is_internal_anchor":false},{"citing_arxiv_id":"2401.15077","citing_title":"EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2412.21187","citing_title":"Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs","ref_index":235,"is_internal_anchor":false},{"citing_arxiv_id":"2404.14469","citing_title":"SnapKV: LLM Knows What You are Looking for Before Generation","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2401.10774","citing_title":"Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads","ref_index":85,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26940","citing_title":"Select to Think: Unlocking SLM Potential with Local Sufficiency","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08632","citing_title":"PARD-2: Target-Aligned Parallel Draft Model for Dual-Mode Speculative Decoding","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09722","citing_title":"ConfigSpec: Profiling-Based Configuration Selection for Distributed Edge--Cloud Speculative LLM Serving","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FIOEL4S7APCV3BFOPD3TDI2HIY","json":"https://pith.science/pith/FIOEL4S7APCV3BFOPD3TDI2HIY.json","graph_json":"https://pith.science/api/pith-number/FIOEL4S7APCV3BFOPD3TDI2HIY/graph.json","events_json":"https://pith.science/api/pith-number/FIOEL4S7APCV3BFOPD3TDI2HIY/events.json","paper":"https://pith.science/paper/FIOEL4S7"},"agent_actions":{"view_html":"https://pith.science/pith/FIOEL4S7APCV3BFOPD3TDI2HIY","download_json":"https://pith.science/pith/FIOEL4S7APCV3BFOPD3TDI2HIY.json","view_paper":"https://pith.science/paper/FIOEL4S7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.09781&json=true","fetch_graph":"https://pith.science/api/pith-number/FIOEL4S7APCV3BFOPD3TDI2HIY/graph.json","fetch_events":"https://pith.science/api/pith-number/FIOEL4S7APCV3BFOPD3TDI2HIY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FIOEL4S7APCV3BFOPD3TDI2HIY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FIOEL4S7APCV3BFOPD3TDI2HIY/action/storage_attestation","attest_author":"https://pith.science/pith/FIOEL4S7APCV3BFOPD3TDI2HIY/action/author_attestation","sign_citation":"https://pith.science/pith/FIOEL4S7APCV3BFOPD3TDI2HIY/action/citation_signature","submit_replication":"https://pith.science/pith/FIOEL4S7APCV3BFOPD3TDI2HIY/action/replication_record"}},"created_at":"2026-07-05T08:02:43.547575+00:00","updated_at":"2026-07-05T08:02:43.547575+00:00"}