{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XVCBLNRC4CUPWRCDPUCBZX46C4","short_pith_number":"pith:XVCBLNRC","schema_version":"1.0","canonical_sha256":"bd4415b622e0a8fb44437d041cdf9e170ff0de4f7ed6bb085ac512b50d9d7933","source":{"kind":"arxiv","id":"2608.06628","version":1},"attestation_state":"computed","paper":{"title":"Retrofitting Linear Attention into Diffusion Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jaeyeon Kim, Jinha Kim, Younghun Roh","submitted_at":"2026-08-06T22:34:20Z","abstract_excerpt":"Diffusion language models (dLLMs) offer a promising alternative to autoregressive models by accelerating inference through parallel decoding. Recent dLLMs commonly use blockwise semi-autoregressive decoding, generating blocks autoregressively while denoising tokens within each active block in parallel. However, despite KV caching, each denoising step still attends to all previous blocks, repeatedly incurring prefix-attention cost. Motivated by this bottleneck, we ask whether dLLM inference can be further accelerated by linearizing attention over previous blocks. We introduce block-hybrid atten"},"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":"2608.06628","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-06T22:34:20Z","cross_cats_sorted":[],"title_canon_sha256":"4e9c37d6c73e5ba30f567d8bf32b8ff050aaf394036743154d6a777de593e353","abstract_canon_sha256":"7ae3861dc82edc8545aa71838f7a3b42f70b547d4fa74a790aa8ea52423272b7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-10T01:11:31.668739Z","signature_b64":"T0MnTvRdWsjxP3v9ZCjN/i1qGawSbce+uvBG6LykEMmHRhS8MlrkfGxEj9hiDWCNyJyRbfYlhO1/FyWi2B2OBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bd4415b622e0a8fb44437d041cdf9e170ff0de4f7ed6bb085ac512b50d9d7933","last_reissued_at":"2026-08-10T01:11:31.666382Z","signature_status":"signed_v1","first_computed_at":"2026-08-10T01:11:31.666382Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Retrofitting Linear Attention into Diffusion Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jaeyeon Kim, Jinha Kim, Younghun Roh","submitted_at":"2026-08-06T22:34:20Z","abstract_excerpt":"Diffusion language models (dLLMs) offer a promising alternative to autoregressive models by accelerating inference through parallel decoding. Recent dLLMs commonly use blockwise semi-autoregressive decoding, generating blocks autoregressively while denoising tokens within each active block in parallel. However, despite KV caching, each denoising step still attends to all previous blocks, repeatedly incurring prefix-attention cost. Motivated by this bottleneck, we ask whether dLLM inference can be further accelerated by linearizing attention over previous blocks. We introduce block-hybrid atten"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.06628","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/2608.06628/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":"2608.06628","created_at":"2026-08-10T01:11:31.667322+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.06628v1","created_at":"2026-08-10T01:11:31.667322+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.06628","created_at":"2026-08-10T01:11:31.667322+00:00"},{"alias_kind":"pith_short_12","alias_value":"XVCBLNRC4CUP","created_at":"2026-08-10T01:11:31.667322+00:00"},{"alias_kind":"pith_short_16","alias_value":"XVCBLNRC4CUPWRCD","created_at":"2026-08-10T01:11:31.667322+00:00"},{"alias_kind":"pith_short_8","alias_value":"XVCBLNRC","created_at":"2026-08-10T01:11:31.667322+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/XVCBLNRC4CUPWRCDPUCBZX46C4","json":"https://pith.science/pith/XVCBLNRC4CUPWRCDPUCBZX46C4.json","graph_json":"https://pith.science/api/pith-number/XVCBLNRC4CUPWRCDPUCBZX46C4/graph.json","events_json":"https://pith.science/api/pith-number/XVCBLNRC4CUPWRCDPUCBZX46C4/events.json","paper":"https://pith.science/paper/XVCBLNRC"},"agent_actions":{"view_html":"https://pith.science/pith/XVCBLNRC4CUPWRCDPUCBZX46C4","download_json":"https://pith.science/pith/XVCBLNRC4CUPWRCDPUCBZX46C4.json","view_paper":"https://pith.science/paper/XVCBLNRC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.06628&json=true","fetch_graph":"https://pith.science/api/pith-number/XVCBLNRC4CUPWRCDPUCBZX46C4/graph.json","fetch_events":"https://pith.science/api/pith-number/XVCBLNRC4CUPWRCDPUCBZX46C4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XVCBLNRC4CUPWRCDPUCBZX46C4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XVCBLNRC4CUPWRCDPUCBZX46C4/action/storage_attestation","attest_author":"https://pith.science/pith/XVCBLNRC4CUPWRCDPUCBZX46C4/action/author_attestation","sign_citation":"https://pith.science/pith/XVCBLNRC4CUPWRCDPUCBZX46C4/action/citation_signature","submit_replication":"https://pith.science/pith/XVCBLNRC4CUPWRCDPUCBZX46C4/action/replication_record"}},"created_at":"2026-08-10T01:11:31.667322+00:00","updated_at":"2026-08-10T01:11:31.667322+00:00"}