{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AHTBYOKFSJBXGAQCRFP2TP5WZC","short_pith_number":"pith:AHTBYOKF","schema_version":"1.0","canonical_sha256":"01e61c39459243730202895fa9bfb6c8adba3c2bf35d4b9c6288ede46af66422","source":{"kind":"arxiv","id":"2605.00161","version":2},"attestation_state":"computed","paper":{"title":"Consistent Diffusion Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A single consistency objective unifies masked and uniform discrete diffusion while delivering state-of-the-art few-step text generation.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hasan Amin, Ming Yin, Rajiv Khanna, Subhojit Som, Xia Song, Yaser Souri, Yuan Gao","submitted_at":"2026-04-30T19:31:02Z","abstract_excerpt":"Diffusion language models (DLMs) are an attractive alternative to autoregressive models because they promise sublinear-time, parallel generation, yet practical gains remain elusive as high-quality samples still demand hundreds of refinement steps. In continuous domains, consistency training along the probability-flow ODE is a popular recipe to accelerate diffusion. For discrete diffusion, no analogous sample-space ODE exists, making direct adaptation ill-defined. We argue that the right discrete substitute is the exact posterior bridge, the closed-form conditional law linking any two noise lev"},"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":"2605.00161","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-30T19:31:02Z","cross_cats_sorted":[],"title_canon_sha256":"d8bed0e70c2150844e7fdc1210b7d6d552a9fd5ab9538a5e7863c65d2bea9b64","abstract_canon_sha256":"826920806f57072956b1ec4d2f9a94ce83c74969f087deea3dd8ec80c6457aee"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:03:47.806932Z","signature_b64":"KZkf9mbsGtCRZqWHDxFhoCWYVJP9cmOL/4UDVUm4iVqiRunv9IFt78RHJeydy4V+ClljSBqxq1s6zuhHKLrDDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"01e61c39459243730202895fa9bfb6c8adba3c2bf35d4b9c6288ede46af66422","last_reissued_at":"2026-06-02T01:03:47.806481Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:03:47.806481Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Consistent Diffusion Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A single consistency objective unifies masked and uniform discrete diffusion while delivering state-of-the-art few-step text generation.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hasan Amin, Ming Yin, Rajiv Khanna, Subhojit Som, Xia Song, Yaser Souri, Yuan Gao","submitted_at":"2026-04-30T19:31:02Z","abstract_excerpt":"Diffusion language models (DLMs) are an attractive alternative to autoregressive models because they promise sublinear-time, parallel generation, yet practical gains remain elusive as high-quality samples still demand hundreds of refinement steps. In continuous domains, consistency training along the probability-flow ODE is a popular recipe to accelerate diffusion. For discrete diffusion, no analogous sample-space ODE exists, making direct adaptation ill-defined. We argue that the right discrete substitute is the exact posterior bridge, the closed-form conditional law linking any two noise lev"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"CDLM establishes a new state of the art on both conditional and unconditional text-generation, consistently outperforming strong base discrete diffusion models and often even multi-stage distilled baselines across sampling budgets, with the largest gains in the few-step regime.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the exact posterior bridge serves as the natural discrete substitute for the probability-flow ODE and that enforcing path-invariance in expectation across these bridges produces superior denoisers without introducing new failure modes.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CDLM trains denoisers to be path-invariant across stochastic posterior bridges in discrete diffusion, unifying prior methods and achieving new SOTA few-step text generation performance.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A single consistency objective unifies masked and uniform discrete diffusion while delivering state-of-the-art few-step text generation.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0975a0bd4c3569125d1239de55abb26c336afb2d0fd5e9777baeb32d5deef9d0"},"source":{"id":"2605.00161","kind":"arxiv","version":2},"verdict":{"id":"ead87992-3fde-4199-99b2-efa1249a7368","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-09T20:56:31.418642Z","strongest_claim":"CDLM establishes a new state of the art on both conditional and unconditional text-generation, consistently outperforming strong base discrete diffusion models and often even multi-stage distilled baselines across sampling budgets, with the largest gains in the few-step regime.","one_line_summary":"CDLM trains denoisers to be path-invariant across stochastic posterior bridges in discrete diffusion, unifying prior methods and achieving new SOTA few-step text generation performance.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the exact posterior bridge serves as the natural discrete substitute for the probability-flow ODE and that enforcing path-invariance in expectation across these bridges produces superior denoisers without introducing new failure modes.","pith_extraction_headline":"A single consistency objective unifies masked and uniform discrete diffusion while delivering state-of-the-art few-step text generation."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.00161/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T20:37:13.938567Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T18:26:16.581082Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"9392d7e32d0d861e4f5c57879526aa9bf96b47aad5b89cdec67c64a0d45e9226"},"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":"2605.00161","created_at":"2026-06-02T01:03:47.806531+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.00161v2","created_at":"2026-06-02T01:03:47.806531+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.00161","created_at":"2026-06-02T01:03:47.806531+00:00"},{"alias_kind":"pith_short_12","alias_value":"AHTBYOKFSJBX","created_at":"2026-06-02T01:03:47.806531+00:00"},{"alias_kind":"pith_short_16","alias_value":"AHTBYOKFSJBXGAQC","created_at":"2026-06-02T01:03:47.806531+00:00"},{"alias_kind":"pith_short_8","alias_value":"AHTBYOKF","created_at":"2026-06-02T01:03:47.806531+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.08810","citing_title":"Continuous Language Diffusion as a Decoder-Interface Problem","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AHTBYOKFSJBXGAQCRFP2TP5WZC","json":"https://pith.science/pith/AHTBYOKFSJBXGAQCRFP2TP5WZC.json","graph_json":"https://pith.science/api/pith-number/AHTBYOKFSJBXGAQCRFP2TP5WZC/graph.json","events_json":"https://pith.science/api/pith-number/AHTBYOKFSJBXGAQCRFP2TP5WZC/events.json","paper":"https://pith.science/paper/AHTBYOKF"},"agent_actions":{"view_html":"https://pith.science/pith/AHTBYOKFSJBXGAQCRFP2TP5WZC","download_json":"https://pith.science/pith/AHTBYOKFSJBXGAQCRFP2TP5WZC.json","view_paper":"https://pith.science/paper/AHTBYOKF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.00161&json=true","fetch_graph":"https://pith.science/api/pith-number/AHTBYOKFSJBXGAQCRFP2TP5WZC/graph.json","fetch_events":"https://pith.science/api/pith-number/AHTBYOKFSJBXGAQCRFP2TP5WZC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AHTBYOKFSJBXGAQCRFP2TP5WZC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AHTBYOKFSJBXGAQCRFP2TP5WZC/action/storage_attestation","attest_author":"https://pith.science/pith/AHTBYOKFSJBXGAQCRFP2TP5WZC/action/author_attestation","sign_citation":"https://pith.science/pith/AHTBYOKFSJBXGAQCRFP2TP5WZC/action/citation_signature","submit_replication":"https://pith.science/pith/AHTBYOKFSJBXGAQCRFP2TP5WZC/action/replication_record"}},"created_at":"2026-06-02T01:03:47.806531+00:00","updated_at":"2026-06-02T01:03:47.806531+00:00"}