{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PEF32I5IVTUQXPBYUCZG3UCJP2","short_pith_number":"pith:PEF32I5I","schema_version":"1.0","canonical_sha256":"790bbd23a8ace90bbc38a0b26dd0497e8cc9831a5760c0ca7991e26399752eb7","source":{"kind":"arxiv","id":"2605.14305","version":1},"attestation_state":"computed","paper":{"title":"Factorization-Error-Free Discrete Diffusion Language Model via Speculative Decoding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"FeF-DLLM eliminates factorization errors in discrete diffusion language models by replacing independent token predictions with an exact prefix-conditioned factorization of the clean posterior.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hang Yuan, Xun Fang, Yunchen Li, Zhou Yu","submitted_at":"2026-05-14T03:15:25Z","abstract_excerpt":"Discrete diffusion language models improve generation efficiency through parallel token prediction, but standard $X_0$ prediction methods introduce factorization errors by approximating the clean token posterior with independent token-wise distributions. This paper proposes Factorization-Error-Free Discrete Diffusion Language Modeling (FeF-DLLM), which replaces independent clean-token prediction with an exact prefix-conditioned factorization of the clean posterior to better preserve token dependencies. To reduce the sequential cost introduced by prefix conditioning, FeF-DLLM further incorporat"},"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":true},"canonical_record":{"source":{"id":"2605.14305","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-14T03:15:25Z","cross_cats_sorted":[],"title_canon_sha256":"1b752d254629e474560257600ffd2c4644cf0f0198229497f92dc83040d93b40","abstract_canon_sha256":"171bf65d9fc87c334c011d41ec3b1aba7f8636d06328f6b5e6b19f6d1db9d2a8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:10.055766Z","signature_b64":"7iIg90FyU8LorW++WpaFI6odiMEp70z/cBmnhV+kiKdld5MCzz0yWFbL6gcOCnr8PrSHVHF1bgp1mP8dyMFBCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"790bbd23a8ace90bbc38a0b26dd0497e8cc9831a5760c0ca7991e26399752eb7","last_reissued_at":"2026-05-17T23:39:10.055299Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:10.055299Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Factorization-Error-Free Discrete Diffusion Language Model via Speculative Decoding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"FeF-DLLM eliminates factorization errors in discrete diffusion language models by replacing independent token predictions with an exact prefix-conditioned factorization of the clean posterior.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hang Yuan, Xun Fang, Yunchen Li, Zhou Yu","submitted_at":"2026-05-14T03:15:25Z","abstract_excerpt":"Discrete diffusion language models improve generation efficiency through parallel token prediction, but standard $X_0$ prediction methods introduce factorization errors by approximating the clean token posterior with independent token-wise distributions. This paper proposes Factorization-Error-Free Discrete Diffusion Language Modeling (FeF-DLLM), which replaces independent clean-token prediction with an exact prefix-conditioned factorization of the clean posterior to better preserve token dependencies. To reduce the sequential cost introduced by prefix conditioning, FeF-DLLM further incorporat"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Theoretically, we prove that FeF-DLLM generates from the true joint distribution and derive its expected acceleration ratio. Experiments demonstrate an average accuracy improvement of 5.04 percentage points and 3.86× inference speedup.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the prefix-conditioned exact factorization can be realized efficiently via speculative decoding without introducing new approximation errors that undermine the joint-distribution guarantee, and that the theoretical acceleration ratio translates to real-world wall-clock gains under the chosen verification strategy.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FeF-DLLM achieves factorization-error-free generation in discrete diffusion language models via prefix-conditioned posterior factorization and speculative decoding, delivering 5.04 pp higher accuracy and 3.86x faster inference on GSM8K, MATH, HumanEval, and MBPP.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"FeF-DLLM eliminates factorization errors in discrete diffusion language models by replacing independent token predictions with an exact prefix-conditioned factorization of the clean posterior.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"89509912be60226ae743191cddd4211db09a505780e39b866a87c9f0fefc587d"},"source":{"id":"2605.14305","kind":"arxiv","version":1},"verdict":{"id":"f48e28fe-d126-41c4-9a4b-9baeb501da81","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T02:49:04.182737Z","strongest_claim":"Theoretically, we prove that FeF-DLLM generates from the true joint distribution and derive its expected acceleration ratio. Experiments demonstrate an average accuracy improvement of 5.04 percentage points and 3.86× inference speedup.","one_line_summary":"FeF-DLLM achieves factorization-error-free generation in discrete diffusion language models via prefix-conditioned posterior factorization and speculative decoding, delivering 5.04 pp higher accuracy and 3.86x faster inference on GSM8K, MATH, HumanEval, and MBPP.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the prefix-conditioned exact factorization can be realized efficiently via speculative decoding without introducing new approximation errors that undermine the joint-distribution guarantee, and that the theoretical acceleration ratio translates to real-world wall-clock gains under the chosen verification strategy.","pith_extraction_headline":"FeF-DLLM eliminates factorization errors in discrete diffusion language models by replacing independent token predictions with an exact prefix-conditioned factorization of the clean posterior."},"references":{"count":13,"sample":[{"doi":"","year":2024,"title":"Program Synthesis with Large Language Models","work_id":"fd241a05-03b9-4de2-9588-9d77ce176125","ref_index":1,"cited_arxiv_id":"2108.07732","is_internal_anchor":true},{"doi":"","year":null,"title":"LLaDA2.0: Scaling Up Diffusion Language Models to 100B","work_id":"a1b1080d-0a91-44a4-8f70-2bf3e7a27e0b","ref_index":2,"cited_arxiv_id":"2512.15745","is_internal_anchor":true},{"doi":"","year":null,"title":"Self-speculative masked diffusions.arXiv preprint arXiv:2510.03929,","work_id":"4f49f986-8a7d-4bfa-9e0c-33519caaa9a1","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Accelerating Large Language Model Decoding with Speculative Sampling","work_id":"b53b11e5-8613-439d-9d9d-0e2a9090d79f","ref_index":4,"cited_arxiv_id":"2302.01318","is_internal_anchor":true},{"doi":"","year":null,"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","ref_index":5,"cited_arxiv_id":"2107.03374","is_internal_anchor":true}],"resolved_work":13,"snapshot_sha256":"7460e2dbcee1d2061b1d53df61f779ae071d565a66a0e621ab9af7ca18a6a8eb","internal_anchors":8},"formal_canon":{"evidence_count":2,"snapshot_sha256":"06034cbd89f150feb7a98554406af84a2d91b5ada8d018649c867161298a20bb"},"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":"2605.14305","created_at":"2026-05-17T23:39:10.055387+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.14305v1","created_at":"2026-05-17T23:39:10.055387+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14305","created_at":"2026-05-17T23:39:10.055387+00:00"},{"alias_kind":"pith_short_12","alias_value":"PEF32I5IVTUQ","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"PEF32I5IVTUQXPBY","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"PEF32I5I","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PEF32I5IVTUQXPBYUCZG3UCJP2","json":"https://pith.science/pith/PEF32I5IVTUQXPBYUCZG3UCJP2.json","graph_json":"https://pith.science/api/pith-number/PEF32I5IVTUQXPBYUCZG3UCJP2/graph.json","events_json":"https://pith.science/api/pith-number/PEF32I5IVTUQXPBYUCZG3UCJP2/events.json","paper":"https://pith.science/paper/PEF32I5I"},"agent_actions":{"view_html":"https://pith.science/pith/PEF32I5IVTUQXPBYUCZG3UCJP2","download_json":"https://pith.science/pith/PEF32I5IVTUQXPBYUCZG3UCJP2.json","view_paper":"https://pith.science/paper/PEF32I5I","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.14305&json=true","fetch_graph":"https://pith.science/api/pith-number/PEF32I5IVTUQXPBYUCZG3UCJP2/graph.json","fetch_events":"https://pith.science/api/pith-number/PEF32I5IVTUQXPBYUCZG3UCJP2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PEF32I5IVTUQXPBYUCZG3UCJP2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PEF32I5IVTUQXPBYUCZG3UCJP2/action/storage_attestation","attest_author":"https://pith.science/pith/PEF32I5IVTUQXPBYUCZG3UCJP2/action/author_attestation","sign_citation":"https://pith.science/pith/PEF32I5IVTUQXPBYUCZG3UCJP2/action/citation_signature","submit_replication":"https://pith.science/pith/PEF32I5IVTUQXPBYUCZG3UCJP2/action/replication_record"}},"created_at":"2026-05-17T23:39:10.055387+00:00","updated_at":"2026-05-17T23:39:10.055387+00:00"}