{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:7ZPMYBVVHAQSC56SGT62D3IM2Q","short_pith_number":"pith:7ZPMYBVV","schema_version":"1.0","canonical_sha256":"fe5ecc06b538212177d234fda1ed0cd4005812e03225ec08de7b05d69032dcb4","source":{"kind":"arxiv","id":"2406.03853","version":1},"attestation_state":"computed","paper":{"title":"Speculative Decoding via Early-exiting for Faster LLM Inference with Thompson Sampling Control Mechanism","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jiahao Liu, Jingang Wang, Qifan Wang, Xunliang Cai","submitted_at":"2024-06-06T08:40:28Z","abstract_excerpt":"The recent advancements in large language models (LLMs) have been extraordinary, yet the escalating inference costs associated with them present challenges in real-world applications. To address these challenges, we propose a novel approach called Early-exiting Speculative Decoding (EESD) with lossless acceleration. Specifically, EESD utilizes a segment of the LLM to generate draft tokens, incorporating Early-exiting structures after the first N layers. To enhance the quality of draft tokens, a self-distillation method is integrated. This early-exiting design not only reduces deployment and tr"},"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":"2406.03853","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-06T08:40:28Z","cross_cats_sorted":[],"title_canon_sha256":"b117caf6571e7ed0f6dda471f76d65342daa210c4e0baaa0598b0c697bb9723b","abstract_canon_sha256":"8f27c1b8374c45d04e8a80195ebbc9a6051a07eb37179ca614ec92230dc31bcb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:28:16.735376Z","signature_b64":"OWJMhuZu95ypdIduiSMxC1FYxqlUPuGcD0oFPivWZil4PXXgS3cvnyoXkZlX03spWfu5s3DmtPxS0YM+8IkgBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fe5ecc06b538212177d234fda1ed0cd4005812e03225ec08de7b05d69032dcb4","last_reissued_at":"2026-07-05T08:28:16.734895Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:28:16.734895Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Speculative Decoding via Early-exiting for Faster LLM Inference with Thompson Sampling Control Mechanism","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jiahao Liu, Jingang Wang, Qifan Wang, Xunliang Cai","submitted_at":"2024-06-06T08:40:28Z","abstract_excerpt":"The recent advancements in large language models (LLMs) have been extraordinary, yet the escalating inference costs associated with them present challenges in real-world applications. To address these challenges, we propose a novel approach called Early-exiting Speculative Decoding (EESD) with lossless acceleration. Specifically, EESD utilizes a segment of the LLM to generate draft tokens, incorporating Early-exiting structures after the first N layers. To enhance the quality of draft tokens, a self-distillation method is integrated. This early-exiting design not only reduces deployment and tr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.03853","kind":"arxiv","version":1},"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/2406.03853/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":"2406.03853","created_at":"2026-07-05T08:28:16.734951+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.03853v1","created_at":"2026-07-05T08:28:16.734951+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.03853","created_at":"2026-07-05T08:28:16.734951+00:00"},{"alias_kind":"pith_short_12","alias_value":"7ZPMYBVVHAQS","created_at":"2026-07-05T08:28:16.734951+00:00"},{"alias_kind":"pith_short_16","alias_value":"7ZPMYBVVHAQSC56S","created_at":"2026-07-05T08:28:16.734951+00:00"},{"alias_kind":"pith_short_8","alias_value":"7ZPMYBVV","created_at":"2026-07-05T08:28:16.734951+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2507.01449","citing_title":"LogitSpec: Accelerating Retrieval-based Speculative Decoding via Next Next Token Speculation","ref_index":42,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7ZPMYBVVHAQSC56SGT62D3IM2Q","json":"https://pith.science/pith/7ZPMYBVVHAQSC56SGT62D3IM2Q.json","graph_json":"https://pith.science/api/pith-number/7ZPMYBVVHAQSC56SGT62D3IM2Q/graph.json","events_json":"https://pith.science/api/pith-number/7ZPMYBVVHAQSC56SGT62D3IM2Q/events.json","paper":"https://pith.science/paper/7ZPMYBVV"},"agent_actions":{"view_html":"https://pith.science/pith/7ZPMYBVVHAQSC56SGT62D3IM2Q","download_json":"https://pith.science/pith/7ZPMYBVVHAQSC56SGT62D3IM2Q.json","view_paper":"https://pith.science/paper/7ZPMYBVV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.03853&json=true","fetch_graph":"https://pith.science/api/pith-number/7ZPMYBVVHAQSC56SGT62D3IM2Q/graph.json","fetch_events":"https://pith.science/api/pith-number/7ZPMYBVVHAQSC56SGT62D3IM2Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7ZPMYBVVHAQSC56SGT62D3IM2Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7ZPMYBVVHAQSC56SGT62D3IM2Q/action/storage_attestation","attest_author":"https://pith.science/pith/7ZPMYBVVHAQSC56SGT62D3IM2Q/action/author_attestation","sign_citation":"https://pith.science/pith/7ZPMYBVVHAQSC56SGT62D3IM2Q/action/citation_signature","submit_replication":"https://pith.science/pith/7ZPMYBVVHAQSC56SGT62D3IM2Q/action/replication_record"}},"created_at":"2026-07-05T08:28:16.734951+00:00","updated_at":"2026-07-05T08:28:16.734951+00:00"}