{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:5TI4EN4PHCPTVFAMB6YINEZCJQ","short_pith_number":"pith:5TI4EN4P","schema_version":"1.0","canonical_sha256":"ecd1c2378f389f3a940c0fb08693224c2151e591377b3f1924c48e2d614ff02b","source":{"kind":"arxiv","id":"2404.08856","version":1},"attestation_state":"computed","paper":{"title":"On Speculative Decoding for Multimodal Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Christopher Lott, Junyoung Park, Mingu Lee, Mukul Gagrani, Raghavv Goel, Wonseok Jeon","submitted_at":"2024-04-13T00:02:36Z","abstract_excerpt":"Inference with Multimodal Large Language Models (MLLMs) is slow due to their large-language-model backbone which suffers from memory bandwidth bottleneck and generates tokens auto-regressively. In this paper, we explore the application of speculative decoding to enhance the inference efficiency of MLLMs, specifically the LLaVA 7B model. We show that a language-only model can serve as a good draft model for speculative decoding with LLaVA 7B, bypassing the need for image tokens and their associated processing components from the draft model. Our experiments across three different tasks show tha"},"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":"2404.08856","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-04-13T00:02:36Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"651d7290a44a6acf689cd0bb9000d27c44ae43b16fe7a99d9b262f16469fd05c","abstract_canon_sha256":"11f2de9543781af861ffe4355114c43c6c4aaa2d607b1cf976e5f4279b0fb409"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:07:41.190295Z","signature_b64":"66pNYJRcGuU9SRUw2BXmtJbPKsZsEvLaEb2eehw94crv08444cHy//dERFI4lQbdLLzD24cpnIeT5K7HqFvxCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ecd1c2378f389f3a940c0fb08693224c2151e591377b3f1924c48e2d614ff02b","last_reissued_at":"2026-07-05T08:07:41.189815Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:07:41.189815Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On Speculative Decoding for Multimodal Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Christopher Lott, Junyoung Park, Mingu Lee, Mukul Gagrani, Raghavv Goel, Wonseok Jeon","submitted_at":"2024-04-13T00:02:36Z","abstract_excerpt":"Inference with Multimodal Large Language Models (MLLMs) is slow due to their large-language-model backbone which suffers from memory bandwidth bottleneck and generates tokens auto-regressively. In this paper, we explore the application of speculative decoding to enhance the inference efficiency of MLLMs, specifically the LLaVA 7B model. We show that a language-only model can serve as a good draft model for speculative decoding with LLaVA 7B, bypassing the need for image tokens and their associated processing components from the draft model. Our experiments across three different tasks show tha"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.08856","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/2404.08856/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":"2404.08856","created_at":"2026-07-05T08:07:41.189875+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.08856v1","created_at":"2026-07-05T08:07:41.189875+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.08856","created_at":"2026-07-05T08:07:41.189875+00:00"},{"alias_kind":"pith_short_12","alias_value":"5TI4EN4PHCPT","created_at":"2026-07-05T08:07:41.189875+00:00"},{"alias_kind":"pith_short_16","alias_value":"5TI4EN4PHCPTVFAM","created_at":"2026-07-05T08:07:41.189875+00:00"},{"alias_kind":"pith_short_8","alias_value":"5TI4EN4P","created_at":"2026-07-05T08:07:41.189875+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/5TI4EN4PHCPTVFAMB6YINEZCJQ","json":"https://pith.science/pith/5TI4EN4PHCPTVFAMB6YINEZCJQ.json","graph_json":"https://pith.science/api/pith-number/5TI4EN4PHCPTVFAMB6YINEZCJQ/graph.json","events_json":"https://pith.science/api/pith-number/5TI4EN4PHCPTVFAMB6YINEZCJQ/events.json","paper":"https://pith.science/paper/5TI4EN4P"},"agent_actions":{"view_html":"https://pith.science/pith/5TI4EN4PHCPTVFAMB6YINEZCJQ","download_json":"https://pith.science/pith/5TI4EN4PHCPTVFAMB6YINEZCJQ.json","view_paper":"https://pith.science/paper/5TI4EN4P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.08856&json=true","fetch_graph":"https://pith.science/api/pith-number/5TI4EN4PHCPTVFAMB6YINEZCJQ/graph.json","fetch_events":"https://pith.science/api/pith-number/5TI4EN4PHCPTVFAMB6YINEZCJQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5TI4EN4PHCPTVFAMB6YINEZCJQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5TI4EN4PHCPTVFAMB6YINEZCJQ/action/storage_attestation","attest_author":"https://pith.science/pith/5TI4EN4PHCPTVFAMB6YINEZCJQ/action/author_attestation","sign_citation":"https://pith.science/pith/5TI4EN4PHCPTVFAMB6YINEZCJQ/action/citation_signature","submit_replication":"https://pith.science/pith/5TI4EN4PHCPTVFAMB6YINEZCJQ/action/replication_record"}},"created_at":"2026-07-05T08:07:41.189875+00:00","updated_at":"2026-07-05T08:07:41.189875+00:00"}