{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:5Q6M2CXR5VQLWUBDRXKW5CY3JM","short_pith_number":"pith:5Q6M2CXR","schema_version":"1.0","canonical_sha256":"ec3ccd0af1ed60bb50238dd56e8b1b4b209e5847c7c90ae9636daaaea59b9816","source":{"kind":"arxiv","id":"2406.14066","version":3},"attestation_state":"computed","paper":{"title":"TurboSpec: Closed-loop Speculation Control System for Optimizing LLM Serving Goodput","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.PF"],"primary_cat":"cs.AI","authors_text":"Alvin Cheung, Chen Zhang, Hao Zhang, Ion Stoica, Jongseok Park, Kaichao You, Kuntai Du, Langxiang Hu, Woosuk Kwon, Xiangxi Mo, Xiaoxuan Liu, Zhijie Deng, Zhuohan Li","submitted_at":"2024-06-20T07:43:33Z","abstract_excerpt":"Large Language Model (LLM) serving systems batch concurrent user requests to achieve efficient serving. However, in real-world deployments, such inter-request parallelism from batching is often limited by external factors such as low request rates or memory constraints. Recent works focus on intra-request parallelism from speculative decoding as a solution to this problem. Unfortunately, benefits from intra-request parallelism are often fragile, as speculative decoding causes overhead, and speculated tokens may miss. We observe that speculative decoding may degrade LLM serving performance if a"},"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":"2406.14066","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-06-20T07:43:33Z","cross_cats_sorted":["cs.PF"],"title_canon_sha256":"af13fa1a6114ef662c39d743df1d4de1ddf4dbe3c9889a70ff960f4acdff0bc5","abstract_canon_sha256":"118d65789dccefa122cc49d2d345961a1b4418a543b142304d4c9fe55c6a1664"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:43:55.769219Z","signature_b64":"Qk4MCKwNsOtZcgAb3lAnCjk3KrpBZfdbcz4/zkJ+DjtTV1QdrykkrerQU/cQbsKFYsZIMDiV3JDTRpOvJQ2vDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ec3ccd0af1ed60bb50238dd56e8b1b4b209e5847c7c90ae9636daaaea59b9816","last_reissued_at":"2026-07-05T11:43:55.768651Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:43:55.768651Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TurboSpec: Closed-loop Speculation Control System for Optimizing LLM Serving Goodput","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.PF"],"primary_cat":"cs.AI","authors_text":"Alvin Cheung, Chen Zhang, Hao Zhang, Ion Stoica, Jongseok Park, Kaichao You, Kuntai Du, Langxiang Hu, Woosuk Kwon, Xiangxi Mo, Xiaoxuan Liu, Zhijie Deng, Zhuohan Li","submitted_at":"2024-06-20T07:43:33Z","abstract_excerpt":"Large Language Model (LLM) serving systems batch concurrent user requests to achieve efficient serving. However, in real-world deployments, such inter-request parallelism from batching is often limited by external factors such as low request rates or memory constraints. Recent works focus on intra-request parallelism from speculative decoding as a solution to this problem. Unfortunately, benefits from intra-request parallelism are often fragile, as speculative decoding causes overhead, and speculated tokens may miss. We observe that speculative decoding may degrade LLM serving performance if a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.14066","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/2406.14066/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":"2406.14066","created_at":"2026-07-05T11:43:55.768710+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.14066v3","created_at":"2026-07-05T11:43:55.768710+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.14066","created_at":"2026-07-05T11:43:55.768710+00:00"},{"alias_kind":"pith_short_12","alias_value":"5Q6M2CXR5VQL","created_at":"2026-07-05T11:43:55.768710+00:00"},{"alias_kind":"pith_short_16","alias_value":"5Q6M2CXR5VQLWUBD","created_at":"2026-07-05T11:43:55.768710+00:00"},{"alias_kind":"pith_short_8","alias_value":"5Q6M2CXR","created_at":"2026-07-05T11:43:55.768710+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25091","citing_title":"Speculation at a Distance: Where Edge-Cloud Speculative Decoding Actually Pays Off","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08446","citing_title":"Sparrow: Sparse Rollout for Stable and Efficient Long-context RL of Large Language Models","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15051","citing_title":"An Interpretable Latency Model for Speculative Decoding in LLM Serving","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24832","citing_title":"Optimus: Elastic Decoding for Efficient Diffusion LLM Serving","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24328","citing_title":"Speculative Verification: Exploiting Information Gain to Refine Speculative Decoding","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2510.15545","citing_title":"TokenTiming: A Dynamic Alignment Method for Universal Speculative Decoding Model Pairs","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09603","citing_title":"ECHO: Elastic Speculative Decoding with Sparse Gating for High-Concurrency Scenarios","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20919","citing_title":"DiP-SD: Distributed Pipelined Speculative Decoding for Efficient LLM Inference at the Edge","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06914","citing_title":"Regulating Branch Parallelism in LLM Serving","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20503","citing_title":"FASER: Fine-Grained Phase Management for Speculative Decoding in Dynamic LLM Serving","ref_index":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5Q6M2CXR5VQLWUBDRXKW5CY3JM","json":"https://pith.science/pith/5Q6M2CXR5VQLWUBDRXKW5CY3JM.json","graph_json":"https://pith.science/api/pith-number/5Q6M2CXR5VQLWUBDRXKW5CY3JM/graph.json","events_json":"https://pith.science/api/pith-number/5Q6M2CXR5VQLWUBDRXKW5CY3JM/events.json","paper":"https://pith.science/paper/5Q6M2CXR"},"agent_actions":{"view_html":"https://pith.science/pith/5Q6M2CXR5VQLWUBDRXKW5CY3JM","download_json":"https://pith.science/pith/5Q6M2CXR5VQLWUBDRXKW5CY3JM.json","view_paper":"https://pith.science/paper/5Q6M2CXR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.14066&json=true","fetch_graph":"https://pith.science/api/pith-number/5Q6M2CXR5VQLWUBDRXKW5CY3JM/graph.json","fetch_events":"https://pith.science/api/pith-number/5Q6M2CXR5VQLWUBDRXKW5CY3JM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5Q6M2CXR5VQLWUBDRXKW5CY3JM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5Q6M2CXR5VQLWUBDRXKW5CY3JM/action/storage_attestation","attest_author":"https://pith.science/pith/5Q6M2CXR5VQLWUBDRXKW5CY3JM/action/author_attestation","sign_citation":"https://pith.science/pith/5Q6M2CXR5VQLWUBDRXKW5CY3JM/action/citation_signature","submit_replication":"https://pith.science/pith/5Q6M2CXR5VQLWUBDRXKW5CY3JM/action/replication_record"}},"created_at":"2026-07-05T11:43:55.768710+00:00","updated_at":"2026-07-05T11:43:55.768710+00:00"}