{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:7H46JJFKFIYE4I3MGAZJ2MSMDO","short_pith_number":"pith:7H46JJFK","canonical_record":{"source":{"id":"2310.16834","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2023-10-25T17:59:12Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"cc8364068344b26b8d4a79347da0c1ecd07a6e2f8459f061e235efb7e3b843ba","abstract_canon_sha256":"31ed072673dc18bf3fac643d5a37bd43f7e61686de4e5d4f4e17a084b6df4a5c"},"schema_version":"1.0"},"canonical_sha256":"f9f9e4a4aa2a304e236c30329d324c1bbe18a9b7b4e16b1208897d589c48a433","source":{"kind":"arxiv","id":"2310.16834","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2310.16834","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"arxiv_version","alias_value":"2310.16834v3","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.16834","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"pith_short_12","alias_value":"7H46JJFKFIYE","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"pith_short_16","alias_value":"7H46JJFKFIYE4I3M","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"pith_short_8","alias_value":"7H46JJFK","created_at":"2026-07-05T08:28:34Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:7H46JJFKFIYE4I3MGAZJ2MSMDO","target":"record","payload":{"canonical_record":{"source":{"id":"2310.16834","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2023-10-25T17:59:12Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"cc8364068344b26b8d4a79347da0c1ecd07a6e2f8459f061e235efb7e3b843ba","abstract_canon_sha256":"31ed072673dc18bf3fac643d5a37bd43f7e61686de4e5d4f4e17a084b6df4a5c"},"schema_version":"1.0"},"canonical_sha256":"f9f9e4a4aa2a304e236c30329d324c1bbe18a9b7b4e16b1208897d589c48a433","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:28:34.612365Z","signature_b64":"XRvrBeuLfqz5EiXDGF38Jd0FTW3G3+eBgYsNfBlwGrOUhGev1KwjJlyGeLMxKy0FYaGyJZ4TvO4Ois+DZNCRAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f9f9e4a4aa2a304e236c30329d324c1bbe18a9b7b4e16b1208897d589c48a433","last_reissued_at":"2026-07-05T08:28:34.611937Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:28:34.611937Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2310.16834","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:28:34Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+tOKdmVYdqlbwzUHYFN4DqW2hTw/KxeTgKn2G1fA0pMnqDjb8dFPV4AeJGtb4ZPk3kNwoD9d8dWszmqtVtxODw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T02:40:33.996316Z"},"content_sha256":"bd16ebe8719164c764a0625a80e9103ea68c8f388983f46921ff95cf3c91eaa3","schema_version":"1.0","event_id":"sha256:bd16ebe8719164c764a0625a80e9103ea68c8f388983f46921ff95cf3c91eaa3"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:7H46JJFKFIYE4I3MGAZJ2MSMDO","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Score entropy extends score matching to discrete spaces for effective language diffusion models.","cross_cats":["cs.CL","cs.LG"],"primary_cat":"stat.ML","authors_text":"Aaron Lou, Chenlin Meng, Stefano Ermon","submitted_at":"2023-10-25T17:59:12Z","abstract_excerpt":"Despite their groundbreaking performance for many generative modeling tasks, diffusion models have fallen short on discrete data domains such as natural language. Crucially, standard diffusion models rely on the well-established theory of score matching, but efforts to generalize this to discrete structures have not yielded the same empirical gains. In this work, we bridge this gap by proposing score entropy, a novel loss that naturally extends score matching to discrete spaces, integrates seamlessly to build discrete diffusion models, and significantly boosts performance. Experimentally, we t"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"we bridge this gap by proposing score entropy, a novel loss that naturally extends score matching to discrete spaces, integrates seamlessly to build discrete diffusion models, and significantly boosts performance.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the score entropy objective, when minimized, produces models whose generated discrete sequences match the true data distribution without hidden biases from the chosen noise schedule or ratio estimation procedure.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Score entropy loss enables discrete diffusion models (SEDD) that cut perplexity 25-75% versus prior diffusion methods and outperform GPT-2 on language modeling while supporting infilling and compute-quality tradeoffs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Score entropy extends score matching to discrete spaces for effective language diffusion models.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"30c2615f6dceee2fdf87656b7c3c7a88ef919fe4041e3944eb613ed080531ebe"},"source":{"id":"2310.16834","kind":"arxiv","version":3},"verdict":{"id":"5de202ee-be40-4bae-a52d-60eeaac5083e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T02:47:37.324079Z","strongest_claim":"we bridge this gap by proposing score entropy, a novel loss that naturally extends score matching to discrete spaces, integrates seamlessly to build discrete diffusion models, and significantly boosts performance.","one_line_summary":"Score entropy loss enables discrete diffusion models (SEDD) that cut perplexity 25-75% versus prior diffusion methods and outperform GPT-2 on language modeling while supporting infilling and compute-quality tradeoffs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the score entropy objective, when minimized, produces models whose generated discrete sequences match the true data distribution without hidden biases from the chosen noise schedule or ratio estimation procedure.","pith_extraction_headline":"Score entropy extends score matching to discrete spaces for effective language diffusion models."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2310.16834/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":15,"sample":[{"doi":"","year":2020,"title":"Analog bits: Generating discrete data using diffusion models with self-conditioning","work_id":"89278cce-9610-430c-a4e1-9c5915aa5c95","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1137/1.9780898718638","year":2001,"title":"org/CorpusID:23284154","work_id":"80394265-2583-4393-b6eb-aeb94b7778e4","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"org/CorpusID:2352990","work_id":"d231368b-6325-45a1-a5e0-bc1802ca18f3","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Tess: Text-to-text self-conditioned simplex diffusion.arXiv preprint arXiv:2305.08379","work_id":"b2b5ac0d-64d8-4444-87f3-108f4ab09f4b","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Hierarchical Text-Conditional Image Generation with CLIP Latents","work_id":"0c6a768b-70b8-4242-bb0e-459f1008c9fc","ref_index":5,"cited_arxiv_id":"2204.06125","is_internal_anchor":true}],"resolved_work":15,"snapshot_sha256":"96eba566de9c9efbfd78cdc99605c8a70fbf66f7e8eaa15e35d98af7b6c8af2b","internal_anchors":2},"formal_canon":{"evidence_count":3,"snapshot_sha256":"2f5f31ae67b631428cc53d6090d37746e100e25e359624f4e9363a65dc3bed99"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"5de202ee-be40-4bae-a52d-60eeaac5083e"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:28:34Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"QrwgFkncNk/Ni0+z0YU9aB3nu+Pn+53Q1Rqmqa/FNLvMZAnGAEj+JsLuvJuOhqMa0nNu/g5LSAMwN+FT5nSbAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T02:40:33.999879Z"},"content_sha256":"642b3e85c1337cff4df7e2378d6a7d6e082ffc2a7d8b725572eadb11fabf987e","schema_version":"1.0","event_id":"sha256:642b3e85c1337cff4df7e2378d6a7d6e082ffc2a7d8b725572eadb11fabf987e"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/7H46JJFKFIYE4I3MGAZJ2MSMDO/bundle.json","state_url":"https://pith.science/pith/7H46JJFKFIYE4I3MGAZJ2MSMDO/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/7H46JJFKFIYE4I3MGAZJ2MSMDO/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-17T02:40:34Z","links":{"resolver":"https://pith.science/pith/7H46JJFKFIYE4I3MGAZJ2MSMDO","bundle":"https://pith.science/pith/7H46JJFKFIYE4I3MGAZJ2MSMDO/bundle.json","state":"https://pith.science/pith/7H46JJFKFIYE4I3MGAZJ2MSMDO/state.json","well_known_bundle":"https://pith.science/.well-known/pith/7H46JJFKFIYE4I3MGAZJ2MSMDO/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:7H46JJFKFIYE4I3MGAZJ2MSMDO","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"31ed072673dc18bf3fac643d5a37bd43f7e61686de4e5d4f4e17a084b6df4a5c","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2023-10-25T17:59:12Z","title_canon_sha256":"cc8364068344b26b8d4a79347da0c1ecd07a6e2f8459f061e235efb7e3b843ba"},"schema_version":"1.0","source":{"id":"2310.16834","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2310.16834","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"arxiv_version","alias_value":"2310.16834v3","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.16834","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"pith_short_12","alias_value":"7H46JJFKFIYE","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"pith_short_16","alias_value":"7H46JJFKFIYE4I3M","created_at":"2026-07-05T08:28:34Z"},{"alias_kind":"pith_short_8","alias_value":"7H46JJFK","created_at":"2026-07-05T08:28:34Z"}],"graph_snapshots":[{"event_id":"sha256:642b3e85c1337cff4df7e2378d6a7d6e082ffc2a7d8b725572eadb11fabf987e","target":"graph","created_at":"2026-07-05T08:28:34Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"we bridge this gap by proposing score entropy, a novel loss that naturally extends score matching to discrete spaces, integrates seamlessly to build discrete diffusion models, and significantly boosts performance."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the score entropy objective, when minimized, produces models whose generated discrete sequences match the true data distribution without hidden biases from the chosen noise schedule or ratio estimation procedure."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Score entropy loss enables discrete diffusion models (SEDD) that cut perplexity 25-75% versus prior diffusion methods and outperform GPT-2 on language modeling while supporting infilling and compute-quality tradeoffs."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Score entropy extends score matching to discrete spaces for effective language diffusion models."}],"snapshot_sha256":"30c2615f6dceee2fdf87656b7c3c7a88ef919fe4041e3944eb613ed080531ebe"},"formal_canon":{"evidence_count":3,"snapshot_sha256":"2f5f31ae67b631428cc53d6090d37746e100e25e359624f4e9363a65dc3bed99"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2310.16834/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Despite their groundbreaking performance for many generative modeling tasks, diffusion models have fallen short on discrete data domains such as natural language. Crucially, standard diffusion models rely on the well-established theory of score matching, but efforts to generalize this to discrete structures have not yielded the same empirical gains. In this work, we bridge this gap by proposing score entropy, a novel loss that naturally extends score matching to discrete spaces, integrates seamlessly to build discrete diffusion models, and significantly boosts performance. Experimentally, we t","authors_text":"Aaron Lou, Chenlin Meng, Stefano Ermon","cross_cats":["cs.CL","cs.LG"],"headline":"Score entropy extends score matching to discrete spaces for effective language diffusion models.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2023-10-25T17:59:12Z","title":"Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution"},"references":{"count":15,"internal_anchors":2,"resolved_work":15,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Analog bits: Generating discrete data using diffusion models with self-conditioning","work_id":"89278cce-9610-430c-a4e1-9c5915aa5c95","year":2020},{"cited_arxiv_id":"","doi":"10.1137/1.9780898718638","is_internal_anchor":false,"ref_index":2,"title":"org/CorpusID:23284154","work_id":"80394265-2583-4393-b6eb-aeb94b7778e4","year":2001},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"org/CorpusID:2352990","work_id":"d231368b-6325-45a1-a5e0-bc1802ca18f3","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Tess: Text-to-text self-conditioned simplex diffusion.arXiv preprint arXiv:2305.08379","work_id":"b2b5ac0d-64d8-4444-87f3-108f4ab09f4b","year":2022},{"cited_arxiv_id":"2204.06125","doi":"","is_internal_anchor":true,"ref_index":5,"title":"Hierarchical Text-Conditional Image Generation with CLIP Latents","work_id":"0c6a768b-70b8-4242-bb0e-459f1008c9fc","year":2022}],"snapshot_sha256":"96eba566de9c9efbfd78cdc99605c8a70fbf66f7e8eaa15e35d98af7b6c8af2b"},"source":{"id":"2310.16834","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-13T02:47:37.324079Z","id":"5de202ee-be40-4bae-a52d-60eeaac5083e","model_set":{"reader":"grok-4.3"},"one_line_summary":"Score entropy loss enables discrete diffusion models (SEDD) that cut perplexity 25-75% versus prior diffusion methods and outperform GPT-2 on language modeling while supporting infilling and compute-quality tradeoffs.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Score entropy extends score matching to discrete spaces for effective language diffusion models.","strongest_claim":"we bridge this gap by proposing score entropy, a novel loss that naturally extends score matching to discrete spaces, integrates seamlessly to build discrete diffusion models, and significantly boosts performance.","weakest_assumption":"That the score entropy objective, when minimized, produces models whose generated discrete sequences match the true data distribution without hidden biases from the chosen noise schedule or ratio estimation procedure."}},"verdict_id":"5de202ee-be40-4bae-a52d-60eeaac5083e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:bd16ebe8719164c764a0625a80e9103ea68c8f388983f46921ff95cf3c91eaa3","target":"record","created_at":"2026-07-05T08:28:34Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"31ed072673dc18bf3fac643d5a37bd43f7e61686de4e5d4f4e17a084b6df4a5c","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2023-10-25T17:59:12Z","title_canon_sha256":"cc8364068344b26b8d4a79347da0c1ecd07a6e2f8459f061e235efb7e3b843ba"},"schema_version":"1.0","source":{"id":"2310.16834","kind":"arxiv","version":3}},"canonical_sha256":"f9f9e4a4aa2a304e236c30329d324c1bbe18a9b7b4e16b1208897d589c48a433","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f9f9e4a4aa2a304e236c30329d324c1bbe18a9b7b4e16b1208897d589c48a433","first_computed_at":"2026-07-05T08:28:34.611937Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:28:34.611937Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"XRvrBeuLfqz5EiXDGF38Jd0FTW3G3+eBgYsNfBlwGrOUhGev1KwjJlyGeLMxKy0FYaGyJZ4TvO4Ois+DZNCRAA==","signature_status":"signed_v1","signed_at":"2026-07-05T08:28:34.612365Z","signed_message":"canonical_sha256_bytes"},"source_id":"2310.16834","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:bd16ebe8719164c764a0625a80e9103ea68c8f388983f46921ff95cf3c91eaa3","sha256:642b3e85c1337cff4df7e2378d6a7d6e082ffc2a7d8b725572eadb11fabf987e"],"state_sha256":"902183a44a63f11f6db46d8a7c7716ecd97eb92df08e77b088cfb0ddd761be76"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"pWcHKw/C2SStvtpcgI7rbTLOykmYTcXlsCmeFQaSEbCPuONsgEGch6Jeutc1oxmb4VIshQIGkrPhnUTRWKdzBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-17T02:40:34.020722Z","bundle_sha256":"04b672cf4519d88f42f803ac0c27b194b844189ff2533151e447ba4873093fb0"}}