{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:HZCBW22J7OP7GROHB22TCSWEBL","short_pith_number":"pith:HZCBW22J","schema_version":"1.0","canonical_sha256":"3e441b6b49fb9ff345c70eb5314ac40ad893bcd9d4ed894741f926bb7fcbc6db","source":{"kind":"arxiv","id":"2506.03206","version":1},"attestation_state":"computed","paper":{"title":"Out-of-Vocabulary Sampling Boosts Speculative Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"David Harel, Hongyang Zhang, Jonathan Mamou, Nadav Timor, Oren Pereg","submitted_at":"2025-06-02T21:08:02Z","abstract_excerpt":"Speculative decoding relies on fast and accurate drafters. Recent state-of-the-art language models employ larger and larger vocabularies, which significantly slows down drafters. One promising approach to boost the efficiency of speculative decoding is to use drafters with smaller vocabularies. However, existing sampling methods cannot draw out-of-vocabulary tokens, creating a tradeoff between drafters' vocabulary size and acceptance rates. This paper introduces Redistributing Drafter Kernels (RDK), the first out-of-vocabulary sampler that effectively recovers acceptance rates by virtually res"},"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":"2506.03206","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-02T21:08:02Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"dbb7cb01f0b2fb6e9ef7eea2e25a73053fb1eb77e03e0fd0e5f1d1d59e43ac6e","abstract_canon_sha256":"65d7583dc3423fb3c50e2c7e177763f0e4cf08a66e9937b0c8d9f4a8f102b2e5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:15:48.144512Z","signature_b64":"rNOi26/MZVgJhAT4XS/EMhHNDS8ARGY5Dx/Wr7YDTiLNs4usxJQU0ZaQFLZ4l8DGoroBY9dqGJ2uSqkgvvEtBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3e441b6b49fb9ff345c70eb5314ac40ad893bcd9d4ed894741f926bb7fcbc6db","last_reissued_at":"2026-07-05T11:15:48.144056Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:15:48.144056Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Out-of-Vocabulary Sampling Boosts Speculative Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"David Harel, Hongyang Zhang, Jonathan Mamou, Nadav Timor, Oren Pereg","submitted_at":"2025-06-02T21:08:02Z","abstract_excerpt":"Speculative decoding relies on fast and accurate drafters. Recent state-of-the-art language models employ larger and larger vocabularies, which significantly slows down drafters. One promising approach to boost the efficiency of speculative decoding is to use drafters with smaller vocabularies. However, existing sampling methods cannot draw out-of-vocabulary tokens, creating a tradeoff between drafters' vocabulary size and acceptance rates. This paper introduces Redistributing Drafter Kernels (RDK), the first out-of-vocabulary sampler that effectively recovers acceptance rates by virtually res"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.03206","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/2506.03206/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":"2506.03206","created_at":"2026-07-05T11:15:48.144111+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.03206v1","created_at":"2026-07-05T11:15:48.144111+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.03206","created_at":"2026-07-05T11:15:48.144111+00:00"},{"alias_kind":"pith_short_12","alias_value":"HZCBW22J7OP7","created_at":"2026-07-05T11:15:48.144111+00:00"},{"alias_kind":"pith_short_16","alias_value":"HZCBW22J7OP7GROH","created_at":"2026-07-05T11:15:48.144111+00:00"},{"alias_kind":"pith_short_8","alias_value":"HZCBW22J","created_at":"2026-07-05T11:15:48.144111+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.10453","citing_title":"SlimSpec: Low-Rank Draft LM-Head for Accelerated Speculative Decoding","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HZCBW22J7OP7GROHB22TCSWEBL","json":"https://pith.science/pith/HZCBW22J7OP7GROHB22TCSWEBL.json","graph_json":"https://pith.science/api/pith-number/HZCBW22J7OP7GROHB22TCSWEBL/graph.json","events_json":"https://pith.science/api/pith-number/HZCBW22J7OP7GROHB22TCSWEBL/events.json","paper":"https://pith.science/paper/HZCBW22J"},"agent_actions":{"view_html":"https://pith.science/pith/HZCBW22J7OP7GROHB22TCSWEBL","download_json":"https://pith.science/pith/HZCBW22J7OP7GROHB22TCSWEBL.json","view_paper":"https://pith.science/paper/HZCBW22J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.03206&json=true","fetch_graph":"https://pith.science/api/pith-number/HZCBW22J7OP7GROHB22TCSWEBL/graph.json","fetch_events":"https://pith.science/api/pith-number/HZCBW22J7OP7GROHB22TCSWEBL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HZCBW22J7OP7GROHB22TCSWEBL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HZCBW22J7OP7GROHB22TCSWEBL/action/storage_attestation","attest_author":"https://pith.science/pith/HZCBW22J7OP7GROHB22TCSWEBL/action/author_attestation","sign_citation":"https://pith.science/pith/HZCBW22J7OP7GROHB22TCSWEBL/action/citation_signature","submit_replication":"https://pith.science/pith/HZCBW22J7OP7GROHB22TCSWEBL/action/replication_record"}},"created_at":"2026-07-05T11:15:48.144111+00:00","updated_at":"2026-07-05T11:15:48.144111+00:00"}