{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:W56OMABIAVXK6JP5VW4TVOSZAN","short_pith_number":"pith:W56OMABI","schema_version":"1.0","canonical_sha256":"b77ce60028056eaf25fdadb93aba59034fafecc972b712426b0b293ee935581d","source":{"kind":"arxiv","id":"2605.16861","version":1},"attestation_state":"computed","paper":{"title":"Prefix-Adaptive Block Diffusion for Efficient Document Recognition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Prefix-Adaptive Block Diffusion replaces fixed block boundaries with causal prefix denoising and dynamic commitment to fix information conflicts in document parsing.","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Chenyu Liu, Dingwei Zhu, Jiazheng Zhang, Jihua Kang, Jun Long, Kaidi Zhang, Mingxu Chai, Qi Zhang, Ruoyu Chen, Tao Gui, Zhiheng Xi, Ziyu Shen","submitted_at":"2026-05-16T07:50:13Z","abstract_excerpt":"Block Diffusion Models (BDMs) support parallel generation, flexible-length output, and KV caching, making them promising for efficient document parsing. However, existing BDMs bind denoising and cache commitment to fixed block boundaries: parallelism shrinks during intra-block denoising, while generated tokens cannot be cached until the whole block is completed. Moreover, intra-block bidirectional denoising conflicts with inter-block autoregression, creating inconsistent information flow that can challenge structure-sensitive recognition. We propose the Prefix-Adaptive Block Diffusion Model (P"},"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":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.16861","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-16T07:50:13Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"28660fb7e126fe914be3c88185a9fed7deef4c015f65c496a73379146ed6bdc9","abstract_canon_sha256":"3d0e029d0acc37b2b84c0cae23919a1995ea8581737a056556de4ed856f03de8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:26.915104Z","signature_b64":"asojRyjy/oAukOVonHoCfPfNjFgEs0p9aWy3LFNNcNSKd4hzkolfo1C+NED/eqmUDZrUqquelpqf404ibTQmAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b77ce60028056eaf25fdadb93aba59034fafecc972b712426b0b293ee935581d","last_reissued_at":"2026-05-20T00:03:26.914245Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:26.914245Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Prefix-Adaptive Block Diffusion for Efficient Document Recognition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Prefix-Adaptive Block Diffusion replaces fixed block boundaries with causal prefix denoising and dynamic commitment to fix information conflicts in document parsing.","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Chenyu Liu, Dingwei Zhu, Jiazheng Zhang, Jihua Kang, Jun Long, Kaidi Zhang, Mingxu Chai, Qi Zhang, Ruoyu Chen, Tao Gui, Zhiheng Xi, Ziyu Shen","submitted_at":"2026-05-16T07:50:13Z","abstract_excerpt":"Block Diffusion Models (BDMs) support parallel generation, flexible-length output, and KV caching, making them promising for efficient document parsing. However, existing BDMs bind denoising and cache commitment to fixed block boundaries: parallelism shrinks during intra-block denoising, while generated tokens cannot be cached until the whole block is completed. Moreover, intra-block bidirectional denoising conflicts with inter-block autoregression, creating inconsistent information flow that can challenge structure-sensitive recognition. We propose the Prefix-Adaptive Block Diffusion Model (P"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The 3B PA-BDM achieves higher recognition scores on several benchmarks and improves inference throughput by 71.6% over the 2.5B MinerU-Diffusion.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That switching to causal prefix-to-suffix denoising inside blocks plus Confidence-gated Structural Loss and Progressive Prefix Commitment will eliminate the information-flow conflict between intra-block and inter-block generation without introducing new errors in structure-sensitive recognition.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"PA-BDM adapts block diffusion by switching to causal intra-block denoising and dynamically committing reliable prefixes to KV cache, yielding higher accuracy and 71.6% higher throughput than a comparable baseline on document benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Prefix-Adaptive Block Diffusion replaces fixed block boundaries with causal prefix denoising and dynamic commitment to fix information conflicts in document parsing.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"569bd622c8b621d00b2012e14a14eaeba77c5bbf920b3711ed615f9f6c41dce6"},"source":{"id":"2605.16861","kind":"arxiv","version":1},"verdict":{"id":"2ce045a0-9869-488a-81f8-b45280242c59","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T20:31:23.947249Z","strongest_claim":"The 3B PA-BDM achieves higher recognition scores on several benchmarks and improves inference throughput by 71.6% over the 2.5B MinerU-Diffusion.","one_line_summary":"PA-BDM adapts block diffusion by switching to causal intra-block denoising and dynamically committing reliable prefixes to KV cache, yielding higher accuracy and 71.6% higher throughput than a comparable baseline on document benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That switching to causal prefix-to-suffix denoising inside blocks plus Confidence-gated Structural Loss and Progressive Prefix Commitment will eliminate the information-flow conflict between intra-block and inter-block generation without introducing new errors in structure-sensitive recognition.","pith_extraction_headline":"Prefix-Adaptive Block Diffusion replaces fixed block boundaries with causal prefix denoising and dynamic commitment to fix information conflicts in document parsing."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.16861/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T21:01:19.229471Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T20:40:55.540745Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T18:41:56.305424Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.380790Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"97720cfa8836f7e171c89f747d23037fd7c6fd90751b1f26fa7962a67c76e64e"},"references":{"count":118,"sample":[{"doi":"","year":1972,"title":"Aho and Jeffrey D","work_id":"b1f5cb43-a3c7-4ea0-85e7-9ccc9dfe1588","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1983,"title":"Publications Manual , year = \"1983\", publisher =","work_id":"aca2b566-99e0-4ebb-9c7a-a81219531259","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1145/322234.322243","year":1981,"title":"Chandra and Dexter C","work_id":"c3270592-bd69-4213-95e1-4aaf8312be9b","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Scalable training of","work_id":"aef70eae-f816-4598-84ec-429a2c09f5fc","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1997,"title":"Dan Gusfield , title =. 1997","work_id":"852d89f5-1e7b-4296-b4f2-71e578b5e9f6","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":118,"snapshot_sha256":"2ad51c5892c48c20cf19312ae994f1c00872b6c2897ca4246bf03eceee6b4486","internal_anchors":1},"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":"2605.16861","created_at":"2026-05-20T00:03:26.914393+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.16861v1","created_at":"2026-05-20T00:03:26.914393+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.16861","created_at":"2026-05-20T00:03:26.914393+00:00"},{"alias_kind":"pith_short_12","alias_value":"W56OMABIAVXK","created_at":"2026-05-20T00:03:26.914393+00:00"},{"alias_kind":"pith_short_16","alias_value":"W56OMABIAVXK6JP5","created_at":"2026-05-20T00:03:26.914393+00:00"},{"alias_kind":"pith_short_8","alias_value":"W56OMABI","created_at":"2026-05-20T00:03:26.914393+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/W56OMABIAVXK6JP5VW4TVOSZAN","json":"https://pith.science/pith/W56OMABIAVXK6JP5VW4TVOSZAN.json","graph_json":"https://pith.science/api/pith-number/W56OMABIAVXK6JP5VW4TVOSZAN/graph.json","events_json":"https://pith.science/api/pith-number/W56OMABIAVXK6JP5VW4TVOSZAN/events.json","paper":"https://pith.science/paper/W56OMABI"},"agent_actions":{"view_html":"https://pith.science/pith/W56OMABIAVXK6JP5VW4TVOSZAN","download_json":"https://pith.science/pith/W56OMABIAVXK6JP5VW4TVOSZAN.json","view_paper":"https://pith.science/paper/W56OMABI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.16861&json=true","fetch_graph":"https://pith.science/api/pith-number/W56OMABIAVXK6JP5VW4TVOSZAN/graph.json","fetch_events":"https://pith.science/api/pith-number/W56OMABIAVXK6JP5VW4TVOSZAN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/W56OMABIAVXK6JP5VW4TVOSZAN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/W56OMABIAVXK6JP5VW4TVOSZAN/action/storage_attestation","attest_author":"https://pith.science/pith/W56OMABIAVXK6JP5VW4TVOSZAN/action/author_attestation","sign_citation":"https://pith.science/pith/W56OMABIAVXK6JP5VW4TVOSZAN/action/citation_signature","submit_replication":"https://pith.science/pith/W56OMABIAVXK6JP5VW4TVOSZAN/action/replication_record"}},"created_at":"2026-05-20T00:03:26.914393+00:00","updated_at":"2026-05-20T00:03:26.914393+00:00"}