{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:2EOS4XE4FQFQ6ZM52UJNWBDG32","short_pith_number":"pith:2EOS4XE4","schema_version":"1.0","canonical_sha256":"d11d2e5c9c2c0b0f659dd512db0466de82c90fafa46c87562614d309af77b647","source":{"kind":"arxiv","id":"2310.08442","version":3},"attestation_state":"computed","paper":{"title":"Unmasking Bias in Diffusion Model Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Feng Zhao, Hongsheng Li, Hu Yu, Jie Huang, Li Shen","submitted_at":"2023-10-12T16:04:41Z","abstract_excerpt":"Denoising diffusion models have emerged as a dominant approach for image generation, however they still suffer from slow convergence in training and color shift issues in sampling. In this paper, we identify that these obstacles can be largely attributed to bias and suboptimality inherent in the default training paradigm of diffusion models. Specifically, we offer theoretical insights that the prevailing constant loss weight strategy in $\\epsilon$-prediction of diffusion models leads to biased estimation during the training phase, hindering accurate estimations of original images. To address t"},"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":"2310.08442","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2023-10-12T16:04:41Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"05abc8faae80b5012148ac6b7761b855d601782b66d1e0371c95ed340b083244","abstract_canon_sha256":"d89eeea0a1b291cbb1fda0a82e8d5bd67adf198b49f6dd232501e521bcd7544a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:51:43.979678Z","signature_b64":"ET78xUtA6feuzwV+GjNnGxbL6Kbp8gwj51/MYfL5lDQQaeENiyJL9LNiCVgK3gutvrYnPkgMd6P6A3IIygzwAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d11d2e5c9c2c0b0f659dd512db0466de82c90fafa46c87562614d309af77b647","last_reissued_at":"2026-07-05T08:51:43.979100Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:51:43.979100Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unmasking Bias in Diffusion Model Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Feng Zhao, Hongsheng Li, Hu Yu, Jie Huang, Li Shen","submitted_at":"2023-10-12T16:04:41Z","abstract_excerpt":"Denoising diffusion models have emerged as a dominant approach for image generation, however they still suffer from slow convergence in training and color shift issues in sampling. In this paper, we identify that these obstacles can be largely attributed to bias and suboptimality inherent in the default training paradigm of diffusion models. Specifically, we offer theoretical insights that the prevailing constant loss weight strategy in $\\epsilon$-prediction of diffusion models leads to biased estimation during the training phase, hindering accurate estimations of original images. To address t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.08442","kind":"arxiv","version":3},"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/2310.08442/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":"2310.08442","created_at":"2026-07-05T08:51:43.979157+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.08442v3","created_at":"2026-07-05T08:51:43.979157+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.08442","created_at":"2026-07-05T08:51:43.979157+00:00"},{"alias_kind":"pith_short_12","alias_value":"2EOS4XE4FQFQ","created_at":"2026-07-05T08:51:43.979157+00:00"},{"alias_kind":"pith_short_16","alias_value":"2EOS4XE4FQFQ6ZM5","created_at":"2026-07-05T08:51:43.979157+00:00"},{"alias_kind":"pith_short_8","alias_value":"2EOS4XE4","created_at":"2026-07-05T08:51:43.979157+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.10790","citing_title":"Elucidating Representation Degradation Problem in Diffusion Model Training","ref_index":61,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2EOS4XE4FQFQ6ZM52UJNWBDG32","json":"https://pith.science/pith/2EOS4XE4FQFQ6ZM52UJNWBDG32.json","graph_json":"https://pith.science/api/pith-number/2EOS4XE4FQFQ6ZM52UJNWBDG32/graph.json","events_json":"https://pith.science/api/pith-number/2EOS4XE4FQFQ6ZM52UJNWBDG32/events.json","paper":"https://pith.science/paper/2EOS4XE4"},"agent_actions":{"view_html":"https://pith.science/pith/2EOS4XE4FQFQ6ZM52UJNWBDG32","download_json":"https://pith.science/pith/2EOS4XE4FQFQ6ZM52UJNWBDG32.json","view_paper":"https://pith.science/paper/2EOS4XE4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.08442&json=true","fetch_graph":"https://pith.science/api/pith-number/2EOS4XE4FQFQ6ZM52UJNWBDG32/graph.json","fetch_events":"https://pith.science/api/pith-number/2EOS4XE4FQFQ6ZM52UJNWBDG32/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2EOS4XE4FQFQ6ZM52UJNWBDG32/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2EOS4XE4FQFQ6ZM52UJNWBDG32/action/storage_attestation","attest_author":"https://pith.science/pith/2EOS4XE4FQFQ6ZM52UJNWBDG32/action/author_attestation","sign_citation":"https://pith.science/pith/2EOS4XE4FQFQ6ZM52UJNWBDG32/action/citation_signature","submit_replication":"https://pith.science/pith/2EOS4XE4FQFQ6ZM52UJNWBDG32/action/replication_record"}},"created_at":"2026-07-05T08:51:43.979157+00:00","updated_at":"2026-07-05T08:51:43.979157+00:00"}