{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:UFZTFQ53RSODP2456T2EQC5LD5","short_pith_number":"pith:UFZTFQ53","schema_version":"1.0","canonical_sha256":"a17332c3bb8c9c37eb9df4f4480bab1f7d0bbb8c22167009676d94ba6a3557fa","source":{"kind":"arxiv","id":"2308.15321","version":6},"attestation_state":"computed","paper":{"title":"Elucidating the Exposure Bias in Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Albert Ali Salah, Itir Onal Ertugrul, Jianlin Su, Mang Ning, Mingxiao Li","submitted_at":"2023-08-29T14:16:09Z","abstract_excerpt":"Diffusion models have demonstrated impressive generative capabilities, but their \\textit{exposure bias} problem, described as the input mismatch between training and sampling, lacks in-depth exploration. In this paper, we systematically investigate the exposure bias problem in diffusion models by first analytically modelling the sampling distribution, based on which we then attribute the prediction error at each sampling step as the root cause of the exposure bias issue. Furthermore, we discuss potential solutions to this issue and propose an intuitive metric for it. Along with the elucidation"},"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":"2308.15321","kind":"arxiv","version":6},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-08-29T14:16:09Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"f30a439d5b473dd2bd55b3e11cbdd551a79538cb5f2d4764d37959c2e42dddee","abstract_canon_sha256":"2e752697a79f9bad200a9d3f0f6f8546b3c0340465a3153bcb65a3c7b304877a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:06:42.511066Z","signature_b64":"pWqthT6seQXH6UnVrKuPKxhDSqJnRNjm2vanjS2R0X2oYZpE5Jui97febAOQ39xDrtBZzXkiVFATzjdQd/8+BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a17332c3bb8c9c37eb9df4f4480bab1f7d0bbb8c22167009676d94ba6a3557fa","last_reissued_at":"2026-07-05T08:06:42.510579Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:06:42.510579Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Elucidating the Exposure Bias in Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Albert Ali Salah, Itir Onal Ertugrul, Jianlin Su, Mang Ning, Mingxiao Li","submitted_at":"2023-08-29T14:16:09Z","abstract_excerpt":"Diffusion models have demonstrated impressive generative capabilities, but their \\textit{exposure bias} problem, described as the input mismatch between training and sampling, lacks in-depth exploration. In this paper, we systematically investigate the exposure bias problem in diffusion models by first analytically modelling the sampling distribution, based on which we then attribute the prediction error at each sampling step as the root cause of the exposure bias issue. Furthermore, we discuss potential solutions to this issue and propose an intuitive metric for it. Along with the elucidation"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2308.15321","kind":"arxiv","version":6},"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/2308.15321/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":"2308.15321","created_at":"2026-07-05T08:06:42.510635+00:00"},{"alias_kind":"arxiv_version","alias_value":"2308.15321v6","created_at":"2026-07-05T08:06:42.510635+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2308.15321","created_at":"2026-07-05T08:06:42.510635+00:00"},{"alias_kind":"pith_short_12","alias_value":"UFZTFQ53RSOD","created_at":"2026-07-05T08:06:42.510635+00:00"},{"alias_kind":"pith_short_16","alias_value":"UFZTFQ53RSODP245","created_at":"2026-07-05T08:06:42.510635+00:00"},{"alias_kind":"pith_short_8","alias_value":"UFZTFQ53","created_at":"2026-07-05T08:06:42.510635+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19162","citing_title":"The Reward Was in Your Data All Along: Correcting Flow Matching with Discriminator-Guided RL","ref_index":137,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31314","citing_title":"AR Forcing: Towards Long-Horizon Robot Navigation World Model","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2411.05472","citing_title":"Bridging the Gap between Learning and Inference for Diffusion-Based Molecule Generation","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20299","citing_title":"Mechanisms of Misgeneralization in Physical Sequence Modeling","ref_index":88,"is_internal_anchor":false},{"citing_arxiv_id":"2602.07775","citing_title":"Rolling Sink: Bridging Limited-Horizon Training and Open-Ended Testing in Autoregressive Video Diffusion","ref_index":69,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UFZTFQ53RSODP2456T2EQC5LD5","json":"https://pith.science/pith/UFZTFQ53RSODP2456T2EQC5LD5.json","graph_json":"https://pith.science/api/pith-number/UFZTFQ53RSODP2456T2EQC5LD5/graph.json","events_json":"https://pith.science/api/pith-number/UFZTFQ53RSODP2456T2EQC5LD5/events.json","paper":"https://pith.science/paper/UFZTFQ53"},"agent_actions":{"view_html":"https://pith.science/pith/UFZTFQ53RSODP2456T2EQC5LD5","download_json":"https://pith.science/pith/UFZTFQ53RSODP2456T2EQC5LD5.json","view_paper":"https://pith.science/paper/UFZTFQ53","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2308.15321&json=true","fetch_graph":"https://pith.science/api/pith-number/UFZTFQ53RSODP2456T2EQC5LD5/graph.json","fetch_events":"https://pith.science/api/pith-number/UFZTFQ53RSODP2456T2EQC5LD5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UFZTFQ53RSODP2456T2EQC5LD5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UFZTFQ53RSODP2456T2EQC5LD5/action/storage_attestation","attest_author":"https://pith.science/pith/UFZTFQ53RSODP2456T2EQC5LD5/action/author_attestation","sign_citation":"https://pith.science/pith/UFZTFQ53RSODP2456T2EQC5LD5/action/citation_signature","submit_replication":"https://pith.science/pith/UFZTFQ53RSODP2456T2EQC5LD5/action/replication_record"}},"created_at":"2026-07-05T08:06:42.510635+00:00","updated_at":"2026-07-05T08:06:42.510635+00:00"}