{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:MRQL27JINYDR3IBAPRUXND4XGV","short_pith_number":"pith:MRQL27JI","schema_version":"1.0","canonical_sha256":"6460bd7d286e071da0207c69768f973561f9c1a5abb5733d366e1e5450bc28d8","source":{"kind":"arxiv","id":"2504.08838","version":2},"attestation_state":"computed","paper":{"title":"SD$^2$: Self-Distilled Sparse Drafters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Mike Lasby, Nish Sinnadurai, Sean Lie, Valavan Manohararajah, Vithursan Thangarasa, Yani Ioannou","submitted_at":"2025-04-10T18:21:17Z","abstract_excerpt":"Speculative decoding is a powerful technique for reducing the latency of Large Language Models (LLMs), offering a fault-tolerant framework that enables the use of highly compressed draft models. In this work, we introduce Self-Distilled Sparse Drafters (SD$^2$), a novel methodology that leverages self-data distillation and fine-grained weight sparsity to produce highly efficient and well-aligned draft models. SD$^2$ systematically enhances draft token acceptance rates while significantly reducing Multiply-Accumulate operations (MACs), even in the Universal Assisted Generation (UAG) setting, wh"},"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":"2504.08838","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-10T18:21:17Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e5c548a8a7b88eab801900742d73a90daacc68c37788169bc4cb5ffbae6fb6d8","abstract_canon_sha256":"becfa520b51363eb181101616bbe57e196a93f6c51bd6dd87dc25c3a6953bbc7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:23.937567Z","signature_b64":"BU6rS8BsvRqTiwZe8Yr7K3PwfEjo9aHnDtxZtovc9Yzvvf/p2+OGH2B4BWEr0JTtYGuWdLD7BH79mvoTLoVQCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6460bd7d286e071da0207c69768f973561f9c1a5abb5733d366e1e5450bc28d8","last_reissued_at":"2026-07-05T11:13:23.937044Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:23.937044Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SD$^2$: Self-Distilled Sparse Drafters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Mike Lasby, Nish Sinnadurai, Sean Lie, Valavan Manohararajah, Vithursan Thangarasa, Yani Ioannou","submitted_at":"2025-04-10T18:21:17Z","abstract_excerpt":"Speculative decoding is a powerful technique for reducing the latency of Large Language Models (LLMs), offering a fault-tolerant framework that enables the use of highly compressed draft models. In this work, we introduce Self-Distilled Sparse Drafters (SD$^2$), a novel methodology that leverages self-data distillation and fine-grained weight sparsity to produce highly efficient and well-aligned draft models. SD$^2$ systematically enhances draft token acceptance rates while significantly reducing Multiply-Accumulate operations (MACs), even in the Universal Assisted Generation (UAG) setting, wh"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.08838","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/2504.08838/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":"2504.08838","created_at":"2026-07-05T11:13:23.937109+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.08838v2","created_at":"2026-07-05T11:13:23.937109+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.08838","created_at":"2026-07-05T11:13:23.937109+00:00"},{"alias_kind":"pith_short_12","alias_value":"MRQL27JINYDR","created_at":"2026-07-05T11:13:23.937109+00:00"},{"alias_kind":"pith_short_16","alias_value":"MRQL27JINYDR3IBA","created_at":"2026-07-05T11:13:23.937109+00:00"},{"alias_kind":"pith_short_8","alias_value":"MRQL27JI","created_at":"2026-07-05T11:13:23.937109+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MRQL27JINYDR3IBAPRUXND4XGV","json":"https://pith.science/pith/MRQL27JINYDR3IBAPRUXND4XGV.json","graph_json":"https://pith.science/api/pith-number/MRQL27JINYDR3IBAPRUXND4XGV/graph.json","events_json":"https://pith.science/api/pith-number/MRQL27JINYDR3IBAPRUXND4XGV/events.json","paper":"https://pith.science/paper/MRQL27JI"},"agent_actions":{"view_html":"https://pith.science/pith/MRQL27JINYDR3IBAPRUXND4XGV","download_json":"https://pith.science/pith/MRQL27JINYDR3IBAPRUXND4XGV.json","view_paper":"https://pith.science/paper/MRQL27JI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.08838&json=true","fetch_graph":"https://pith.science/api/pith-number/MRQL27JINYDR3IBAPRUXND4XGV/graph.json","fetch_events":"https://pith.science/api/pith-number/MRQL27JINYDR3IBAPRUXND4XGV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MRQL27JINYDR3IBAPRUXND4XGV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MRQL27JINYDR3IBAPRUXND4XGV/action/storage_attestation","attest_author":"https://pith.science/pith/MRQL27JINYDR3IBAPRUXND4XGV/action/author_attestation","sign_citation":"https://pith.science/pith/MRQL27JINYDR3IBAPRUXND4XGV/action/citation_signature","submit_replication":"https://pith.science/pith/MRQL27JINYDR3IBAPRUXND4XGV/action/replication_record"}},"created_at":"2026-07-05T11:13:23.937109+00:00","updated_at":"2026-07-05T11:13:23.937109+00:00"}