{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:Y52HCMJ7CC6X5BGV4QABWQBAF2","short_pith_number":"pith:Y52HCMJ7","schema_version":"1.0","canonical_sha256":"c77471313f10bd7e84d5e4001b40202e81ab0ac361a34d13740bee791c94ea30","source":{"kind":"arxiv","id":"2506.01206","version":1},"attestation_state":"computed","paper":{"title":"Mamba Drafters for Speculative Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Aram Galstyan, Daewon Choi, Insu Han, Jihoon Tack, Jinwoo Shin, Kyuyoung Kim, Saket Dingliwal, Seojin Kim, Seunghyuk Oh, Shubham Katiyar, Sravan Babu Bodapati, Woomin Song","submitted_at":"2025-06-01T22:52:47Z","abstract_excerpt":"Speculative decoding has emerged as a promising approach to accelerating large language model (LLM) generation using a fast drafter while maintaining alignment with the target model's distribution. However, existing approaches face a trade-off: external drafters offer flexibility but can suffer from slower drafting, while self-speculation methods use drafters tailored to the target model but require re-training. In this paper, we introduce novel drafters based on Mamba, a state-of-the-art state space model (SSM), as a solution that combines the best aspects of both approaches. By leveraging th"},"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":"2506.01206","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-01T22:52:47Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f00870ffd05fb74984ec493fb2cd42b3abc334ce06a293afc900946fb561bb5e","abstract_canon_sha256":"195abeb3ab90bc14858eafb8fa2a30c6ab503093796794665c256e25f6ad4de9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:48.128533Z","signature_b64":"6bsS/Mv13jJHU52KODLYQ8n4R0VWamIaq6kscackiBvGM9q5l9iyQixTyU874g0jjLa2frCI9mdaYZ/WJ6bdCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c77471313f10bd7e84d5e4001b40202e81ab0ac361a34d13740bee791c94ea30","last_reissued_at":"2026-07-05T11:13:48.128071Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:48.128071Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mamba Drafters for Speculative Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Aram Galstyan, Daewon Choi, Insu Han, Jihoon Tack, Jinwoo Shin, Kyuyoung Kim, Saket Dingliwal, Seojin Kim, Seunghyuk Oh, Shubham Katiyar, Sravan Babu Bodapati, Woomin Song","submitted_at":"2025-06-01T22:52:47Z","abstract_excerpt":"Speculative decoding has emerged as a promising approach to accelerating large language model (LLM) generation using a fast drafter while maintaining alignment with the target model's distribution. However, existing approaches face a trade-off: external drafters offer flexibility but can suffer from slower drafting, while self-speculation methods use drafters tailored to the target model but require re-training. In this paper, we introduce novel drafters based on Mamba, a state-of-the-art state space model (SSM), as a solution that combines the best aspects of both approaches. By leveraging th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.01206","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/2506.01206/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":"2506.01206","created_at":"2026-07-05T11:13:48.128126+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.01206v1","created_at":"2026-07-05T11:13:48.128126+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.01206","created_at":"2026-07-05T11:13:48.128126+00:00"},{"alias_kind":"pith_short_12","alias_value":"Y52HCMJ7CC6X","created_at":"2026-07-05T11:13:48.128126+00:00"},{"alias_kind":"pith_short_16","alias_value":"Y52HCMJ7CC6X5BGV","created_at":"2026-07-05T11:13:48.128126+00:00"},{"alias_kind":"pith_short_8","alias_value":"Y52HCMJ7","created_at":"2026-07-05T11:13:48.128126+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.01106","citing_title":"Component-Aware Self-Speculative Decoding in Hybrid Language Models","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Y52HCMJ7CC6X5BGV4QABWQBAF2","json":"https://pith.science/pith/Y52HCMJ7CC6X5BGV4QABWQBAF2.json","graph_json":"https://pith.science/api/pith-number/Y52HCMJ7CC6X5BGV4QABWQBAF2/graph.json","events_json":"https://pith.science/api/pith-number/Y52HCMJ7CC6X5BGV4QABWQBAF2/events.json","paper":"https://pith.science/paper/Y52HCMJ7"},"agent_actions":{"view_html":"https://pith.science/pith/Y52HCMJ7CC6X5BGV4QABWQBAF2","download_json":"https://pith.science/pith/Y52HCMJ7CC6X5BGV4QABWQBAF2.json","view_paper":"https://pith.science/paper/Y52HCMJ7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.01206&json=true","fetch_graph":"https://pith.science/api/pith-number/Y52HCMJ7CC6X5BGV4QABWQBAF2/graph.json","fetch_events":"https://pith.science/api/pith-number/Y52HCMJ7CC6X5BGV4QABWQBAF2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Y52HCMJ7CC6X5BGV4QABWQBAF2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Y52HCMJ7CC6X5BGV4QABWQBAF2/action/storage_attestation","attest_author":"https://pith.science/pith/Y52HCMJ7CC6X5BGV4QABWQBAF2/action/author_attestation","sign_citation":"https://pith.science/pith/Y52HCMJ7CC6X5BGV4QABWQBAF2/action/citation_signature","submit_replication":"https://pith.science/pith/Y52HCMJ7CC6X5BGV4QABWQBAF2/action/replication_record"}},"created_at":"2026-07-05T11:13:48.128126+00:00","updated_at":"2026-07-05T11:13:48.128126+00:00"}