{"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"}