{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:J7VL6VSDKR237AVT7KDRAJDVGY","short_pith_number":"pith:J7VL6VSD","schema_version":"1.0","canonical_sha256":"4feabf56435475bf82b3fa87102475360f086865ded0d2cb31fbae02774c4b9a","source":{"kind":"arxiv","id":"2502.05609","version":1},"attestation_state":"computed","paper":{"title":"Lossless Acceleration of Large Language Models with Hierarchical Drafting based on Temporal Locality in Speculative Decoding","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hoyun Song, Huije Lee, Jeongyeon Seo, Jong C. Park, Sangjin Choi, Soyeong Jeong, Sukmin Cho, Taeho Hwang, Youngjin Kwon","submitted_at":"2025-02-08T15:32:53Z","abstract_excerpt":"Accelerating inference in Large Language Models (LLMs) is critical for real-time interactions, as they have been widely incorporated into real-world services. Speculative decoding, a fully algorithmic solution, has gained attention for improving inference speed by drafting and verifying tokens, thereby generating multiple tokens in a single forward pass. However, current drafting strategies usually require significant fine-tuning or have inconsistent performance across tasks. To address these challenges, we propose Hierarchy Drafting (HD), a novel lossless drafting approach that organizes vari"},"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":"2502.05609","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-08T15:32:53Z","cross_cats_sorted":[],"title_canon_sha256":"db090370d19bee1e08fbee4838c7f0190d6c8d4488f27a8daa985614045b0de9","abstract_canon_sha256":"62d17bc59641aa9c51521f39ad33be4fc5cd10d8b64c939af6deafc7e6293c43"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:11:40.498876Z","signature_b64":"J3O3xPy1AI2GTUap6rh8kmi6wPJlswMj8ctut+0F8bjGyLPiSR9VMRl1w+jYmuthaFaUM6eD6Etn1KBBMpH4CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4feabf56435475bf82b3fa87102475360f086865ded0d2cb31fbae02774c4b9a","last_reissued_at":"2026-07-05T10:11:40.498482Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:11:40.498482Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Lossless Acceleration of Large Language Models with Hierarchical Drafting based on Temporal Locality in Speculative Decoding","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hoyun Song, Huije Lee, Jeongyeon Seo, Jong C. Park, Sangjin Choi, Soyeong Jeong, Sukmin Cho, Taeho Hwang, Youngjin Kwon","submitted_at":"2025-02-08T15:32:53Z","abstract_excerpt":"Accelerating inference in Large Language Models (LLMs) is critical for real-time interactions, as they have been widely incorporated into real-world services. Speculative decoding, a fully algorithmic solution, has gained attention for improving inference speed by drafting and verifying tokens, thereby generating multiple tokens in a single forward pass. However, current drafting strategies usually require significant fine-tuning or have inconsistent performance across tasks. To address these challenges, we propose Hierarchy Drafting (HD), a novel lossless drafting approach that organizes vari"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.05609","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/2502.05609/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":"2502.05609","created_at":"2026-07-05T10:11:40.498531+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.05609v1","created_at":"2026-07-05T10:11:40.498531+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.05609","created_at":"2026-07-05T10:11:40.498531+00:00"},{"alias_kind":"pith_short_12","alias_value":"J7VL6VSDKR23","created_at":"2026-07-05T10:11:40.498531+00:00"},{"alias_kind":"pith_short_16","alias_value":"J7VL6VSDKR237AVT","created_at":"2026-07-05T10:11:40.498531+00:00"},{"alias_kind":"pith_short_8","alias_value":"J7VL6VSD","created_at":"2026-07-05T10:11:40.498531+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2603.17573","citing_title":"HeiSD: Hybrid Speculative Decoding for Embodied Vision-Language-Action Models with Kinematic Awareness","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/J7VL6VSDKR237AVT7KDRAJDVGY","json":"https://pith.science/pith/J7VL6VSDKR237AVT7KDRAJDVGY.json","graph_json":"https://pith.science/api/pith-number/J7VL6VSDKR237AVT7KDRAJDVGY/graph.json","events_json":"https://pith.science/api/pith-number/J7VL6VSDKR237AVT7KDRAJDVGY/events.json","paper":"https://pith.science/paper/J7VL6VSD"},"agent_actions":{"view_html":"https://pith.science/pith/J7VL6VSDKR237AVT7KDRAJDVGY","download_json":"https://pith.science/pith/J7VL6VSDKR237AVT7KDRAJDVGY.json","view_paper":"https://pith.science/paper/J7VL6VSD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.05609&json=true","fetch_graph":"https://pith.science/api/pith-number/J7VL6VSDKR237AVT7KDRAJDVGY/graph.json","fetch_events":"https://pith.science/api/pith-number/J7VL6VSDKR237AVT7KDRAJDVGY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/J7VL6VSDKR237AVT7KDRAJDVGY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/J7VL6VSDKR237AVT7KDRAJDVGY/action/storage_attestation","attest_author":"https://pith.science/pith/J7VL6VSDKR237AVT7KDRAJDVGY/action/author_attestation","sign_citation":"https://pith.science/pith/J7VL6VSDKR237AVT7KDRAJDVGY/action/citation_signature","submit_replication":"https://pith.science/pith/J7VL6VSDKR237AVT7KDRAJDVGY/action/replication_record"}},"created_at":"2026-07-05T10:11:40.498531+00:00","updated_at":"2026-07-05T10:11:40.498531+00:00"}