{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:6TG7BLW4FN3U2AKOTHAA2GQOHZ","short_pith_number":"pith:6TG7BLW4","schema_version":"1.0","canonical_sha256":"f4cdf0aedc2b774d014e99c00d1a0e3e7c3d9d5c1a484069d35b7034e2a06a69","source":{"kind":"arxiv","id":"2412.12639","version":3},"attestation_state":"computed","paper":{"title":"Falcon: Faster and Parallel Inference of Large Language Models through Enhanced Semi-Autoregressive Drafting and Custom-Designed Decoding Tree","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Feng Ji, Weisheng Xie, Xiangxiang Gao, Yiwei Xiang","submitted_at":"2024-12-17T08:02:08Z","abstract_excerpt":"Striking an optimal balance between minimal drafting latency and high speculation accuracy to enhance the inference speed of Large Language Models remains a significant challenge in speculative decoding. In this paper, we introduce Falcon, an innovative semi-autoregressive speculative decoding framework fashioned to augment both the drafter's parallelism and output quality. Falcon incorporates the Coupled Sequential Glancing Distillation technique, which fortifies inter-token dependencies within the same block, leading to increased speculation accuracy. We offer a comprehensive theoretical ana"},"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":"2412.12639","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-12-17T08:02:08Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f0d9df24deaa493cc6b704ea7fe3bf8a9e2717f0d235159c9c95807627a69623","abstract_canon_sha256":"cbf387a4961e480c8e207200df3d303793d255091fc1c5996dfa65ac4e665812"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:52:17.144480Z","signature_b64":"MhForKRgg9mBMj6G2W99RNCMAb0dmBNiHqmMOKB4zUAmsZGaHHem5UmLBSEJZ5UdunE8Gbs1Nuwsn1n36GrWCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f4cdf0aedc2b774d014e99c00d1a0e3e7c3d9d5c1a484069d35b7034e2a06a69","last_reissued_at":"2026-07-05T10:52:17.143981Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:52:17.143981Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Falcon: Faster and Parallel Inference of Large Language Models through Enhanced Semi-Autoregressive Drafting and Custom-Designed Decoding Tree","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Feng Ji, Weisheng Xie, Xiangxiang Gao, Yiwei Xiang","submitted_at":"2024-12-17T08:02:08Z","abstract_excerpt":"Striking an optimal balance between minimal drafting latency and high speculation accuracy to enhance the inference speed of Large Language Models remains a significant challenge in speculative decoding. In this paper, we introduce Falcon, an innovative semi-autoregressive speculative decoding framework fashioned to augment both the drafter's parallelism and output quality. Falcon incorporates the Coupled Sequential Glancing Distillation technique, which fortifies inter-token dependencies within the same block, leading to increased speculation accuracy. We offer a comprehensive theoretical ana"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.12639","kind":"arxiv","version":3},"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/2412.12639/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":"2412.12639","created_at":"2026-07-05T10:52:17.144035+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.12639v3","created_at":"2026-07-05T10:52:17.144035+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.12639","created_at":"2026-07-05T10:52:17.144035+00:00"},{"alias_kind":"pith_short_12","alias_value":"6TG7BLW4FN3U","created_at":"2026-07-05T10:52:17.144035+00:00"},{"alias_kind":"pith_short_16","alias_value":"6TG7BLW4FN3U2AKO","created_at":"2026-07-05T10:52:17.144035+00:00"},{"alias_kind":"pith_short_8","alias_value":"6TG7BLW4","created_at":"2026-07-05T10:52:17.144035+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.08632","citing_title":"PARD-2: Target-Aligned Parallel Draft Model for Dual-Mode Speculative Decoding","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6TG7BLW4FN3U2AKOTHAA2GQOHZ","json":"https://pith.science/pith/6TG7BLW4FN3U2AKOTHAA2GQOHZ.json","graph_json":"https://pith.science/api/pith-number/6TG7BLW4FN3U2AKOTHAA2GQOHZ/graph.json","events_json":"https://pith.science/api/pith-number/6TG7BLW4FN3U2AKOTHAA2GQOHZ/events.json","paper":"https://pith.science/paper/6TG7BLW4"},"agent_actions":{"view_html":"https://pith.science/pith/6TG7BLW4FN3U2AKOTHAA2GQOHZ","download_json":"https://pith.science/pith/6TG7BLW4FN3U2AKOTHAA2GQOHZ.json","view_paper":"https://pith.science/paper/6TG7BLW4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.12639&json=true","fetch_graph":"https://pith.science/api/pith-number/6TG7BLW4FN3U2AKOTHAA2GQOHZ/graph.json","fetch_events":"https://pith.science/api/pith-number/6TG7BLW4FN3U2AKOTHAA2GQOHZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6TG7BLW4FN3U2AKOTHAA2GQOHZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6TG7BLW4FN3U2AKOTHAA2GQOHZ/action/storage_attestation","attest_author":"https://pith.science/pith/6TG7BLW4FN3U2AKOTHAA2GQOHZ/action/author_attestation","sign_citation":"https://pith.science/pith/6TG7BLW4FN3U2AKOTHAA2GQOHZ/action/citation_signature","submit_replication":"https://pith.science/pith/6TG7BLW4FN3U2AKOTHAA2GQOHZ/action/replication_record"}},"created_at":"2026-07-05T10:52:17.144035+00:00","updated_at":"2026-07-05T10:52:17.144035+00:00"}