{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XY7LTV3CJ6USE4SJNV3TFDJTEL","short_pith_number":"pith:XY7LTV3C","schema_version":"1.0","canonical_sha256":"be3eb9d7624fa92272496d77328d3322ed6bbd0bf661e5d77e2b43ff10c1ce2c","source":{"kind":"arxiv","id":"2606.18394","version":1},"attestation_state":"computed","paper":{"title":"JetFlow: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bojun Wang, Daxin Jiang, Haoran Yuan, Hao Zhang, Lanxiang Hu, Tajana Rosing, Yibo Zhu, Yujie Zhao, Yulun Wu, Yu-Yang Qian, Zhaoxiang Feng","submitted_at":"2026-06-16T18:37:32Z","abstract_excerpt":"Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it faces a scaling limitation: increasing the draft budget improves speed only when acceptance remains high and drafting overhead stays low. This ceiling has been difficult to break because prior head-based SD methods face a causality-efficiency dilemma. Autoregressive drafters produce path-conditioned candidates that are effective for tree speculative decoding with higher acceptance length, but their drafting cost grows with tree depth. Bidirectiona"},"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":"2606.18394","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-16T18:37:32Z","cross_cats_sorted":[],"title_canon_sha256":"aacd61f22315ee4ff12a93aa015aa6e1faf3ec91a61dfc5226573786e19f30cd","abstract_canon_sha256":"e1bc04870450e8095c4256f1fba671a805e7a3bfed528e1c8172c4d54a99e980"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:11:00.204365Z","signature_b64":"eex1Ez8rtgI0eDJP+IxjdCciCvEsxGmPvDNostm7W6CTev9O7hNcgNxE7QNEOWuArc117uXHHqS/VcdqGAZUAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"be3eb9d7624fa92272496d77328d3322ed6bbd0bf661e5d77e2b43ff10c1ce2c","last_reissued_at":"2026-06-19T16:11:00.204006Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:11:00.204006Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"JetFlow: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bojun Wang, Daxin Jiang, Haoran Yuan, Hao Zhang, Lanxiang Hu, Tajana Rosing, Yibo Zhu, Yujie Zhao, Yulun Wu, Yu-Yang Qian, Zhaoxiang Feng","submitted_at":"2026-06-16T18:37:32Z","abstract_excerpt":"Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it faces a scaling limitation: increasing the draft budget improves speed only when acceptance remains high and drafting overhead stays low. This ceiling has been difficult to break because prior head-based SD methods face a causality-efficiency dilemma. Autoregressive drafters produce path-conditioned candidates that are effective for tree speculative decoding with higher acceptance length, but their drafting cost grows with tree depth. Bidirectiona"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.18394","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/2606.18394/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":"2606.18394","created_at":"2026-06-19T16:11:00.204061+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.18394v1","created_at":"2026-06-19T16:11:00.204061+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.18394","created_at":"2026-06-19T16:11:00.204061+00:00"},{"alias_kind":"pith_short_12","alias_value":"XY7LTV3CJ6US","created_at":"2026-06-19T16:11:00.204061+00:00"},{"alias_kind":"pith_short_16","alias_value":"XY7LTV3CJ6USE4SJ","created_at":"2026-06-19T16:11:00.204061+00:00"},{"alias_kind":"pith_short_8","alias_value":"XY7LTV3C","created_at":"2026-06-19T16:11:00.204061+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/XY7LTV3CJ6USE4SJNV3TFDJTEL","json":"https://pith.science/pith/XY7LTV3CJ6USE4SJNV3TFDJTEL.json","graph_json":"https://pith.science/api/pith-number/XY7LTV3CJ6USE4SJNV3TFDJTEL/graph.json","events_json":"https://pith.science/api/pith-number/XY7LTV3CJ6USE4SJNV3TFDJTEL/events.json","paper":"https://pith.science/paper/XY7LTV3C"},"agent_actions":{"view_html":"https://pith.science/pith/XY7LTV3CJ6USE4SJNV3TFDJTEL","download_json":"https://pith.science/pith/XY7LTV3CJ6USE4SJNV3TFDJTEL.json","view_paper":"https://pith.science/paper/XY7LTV3C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.18394&json=true","fetch_graph":"https://pith.science/api/pith-number/XY7LTV3CJ6USE4SJNV3TFDJTEL/graph.json","fetch_events":"https://pith.science/api/pith-number/XY7LTV3CJ6USE4SJNV3TFDJTEL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XY7LTV3CJ6USE4SJNV3TFDJTEL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XY7LTV3CJ6USE4SJNV3TFDJTEL/action/storage_attestation","attest_author":"https://pith.science/pith/XY7LTV3CJ6USE4SJNV3TFDJTEL/action/author_attestation","sign_citation":"https://pith.science/pith/XY7LTV3CJ6USE4SJNV3TFDJTEL/action/citation_signature","submit_replication":"https://pith.science/pith/XY7LTV3CJ6USE4SJNV3TFDJTEL/action/replication_record"}},"created_at":"2026-06-19T16:11:00.204061+00:00","updated_at":"2026-06-19T16:11:00.204061+00:00"}