{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:A2QJHKSYJH2JFDTBIGC6D7SAXW","short_pith_number":"pith:A2QJHKSY","schema_version":"1.0","canonical_sha256":"06a093aa5849f4928e614185e1fe40bd85ca32d1053a9ce0d1bb3ac2c56debca","source":{"kind":"arxiv","id":"2308.00755","version":2},"attestation_state":"computed","paper":{"title":"The Bias Amplification Paradox in Text-to-Image Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CV","cs.CY"],"primary_cat":"cs.LG","authors_text":"Preethi Seshadri, Sameer Singh, Yanai Elazar","submitted_at":"2023-08-01T18:00:08Z","abstract_excerpt":"Bias amplification is a phenomenon in which models exacerbate biases or stereotypes present in the training data. In this paper, we study bias amplification in the text-to-image domain using Stable Diffusion by comparing gender ratios in training vs. generated images. We find that the model appears to amplify gender-occupation biases found in the training data (LAION) considerably. However, we discover that amplification can be largely attributed to discrepancies between training captions and model prompts. For example, an inherent difference is that captions from the training data often conta"},"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":"2308.00755","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-08-01T18:00:08Z","cross_cats_sorted":["cs.CL","cs.CV","cs.CY"],"title_canon_sha256":"1a3f3fb53043465195b6503c6cf71da9a669481e9f25b7c83ef9804b4ca8bac6","abstract_canon_sha256":"aaad6b062e41b582bcc55fb41f0f03eaf92ec8b4d3ce094b0c21e7839ac2d908"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:12:51.948585Z","signature_b64":"gHhZBVtFXhj0jzpafmGXz1OkDRl3/9fx9VH5G6RcXQdCQt/hBZvnWEl62CXm/6jn098/3Q9O4BQl3HyO4sqbDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"06a093aa5849f4928e614185e1fe40bd85ca32d1053a9ce0d1bb3ac2c56debca","last_reissued_at":"2026-07-05T07:12:51.948147Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:12:51.948147Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Bias Amplification Paradox in Text-to-Image Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CV","cs.CY"],"primary_cat":"cs.LG","authors_text":"Preethi Seshadri, Sameer Singh, Yanai Elazar","submitted_at":"2023-08-01T18:00:08Z","abstract_excerpt":"Bias amplification is a phenomenon in which models exacerbate biases or stereotypes present in the training data. In this paper, we study bias amplification in the text-to-image domain using Stable Diffusion by comparing gender ratios in training vs. generated images. We find that the model appears to amplify gender-occupation biases found in the training data (LAION) considerably. However, we discover that amplification can be largely attributed to discrepancies between training captions and model prompts. For example, an inherent difference is that captions from the training data often conta"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2308.00755","kind":"arxiv","version":2},"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/2308.00755/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":"2308.00755","created_at":"2026-07-05T07:12:51.948202+00:00"},{"alias_kind":"arxiv_version","alias_value":"2308.00755v2","created_at":"2026-07-05T07:12:51.948202+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2308.00755","created_at":"2026-07-05T07:12:51.948202+00:00"},{"alias_kind":"pith_short_12","alias_value":"A2QJHKSYJH2J","created_at":"2026-07-05T07:12:51.948202+00:00"},{"alias_kind":"pith_short_16","alias_value":"A2QJHKSYJH2JFDTB","created_at":"2026-07-05T07:12:51.948202+00:00"},{"alias_kind":"pith_short_8","alias_value":"A2QJHKSY","created_at":"2026-07-05T07:12:51.948202+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.00817","citing_title":"Training-Free Debiasing of Diffusion Models via CLIP-Guided Denoising Optimization","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17118","citing_title":"Differentiable Optimization Layers for Guaranteed Fairness in Deep Learning","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12481","citing_title":"T2I-BiasBench: A Multi-Metric Framework for Auditing Demographic and Cultural Bias in Text-to-Image Models","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/A2QJHKSYJH2JFDTBIGC6D7SAXW","json":"https://pith.science/pith/A2QJHKSYJH2JFDTBIGC6D7SAXW.json","graph_json":"https://pith.science/api/pith-number/A2QJHKSYJH2JFDTBIGC6D7SAXW/graph.json","events_json":"https://pith.science/api/pith-number/A2QJHKSYJH2JFDTBIGC6D7SAXW/events.json","paper":"https://pith.science/paper/A2QJHKSY"},"agent_actions":{"view_html":"https://pith.science/pith/A2QJHKSYJH2JFDTBIGC6D7SAXW","download_json":"https://pith.science/pith/A2QJHKSYJH2JFDTBIGC6D7SAXW.json","view_paper":"https://pith.science/paper/A2QJHKSY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2308.00755&json=true","fetch_graph":"https://pith.science/api/pith-number/A2QJHKSYJH2JFDTBIGC6D7SAXW/graph.json","fetch_events":"https://pith.science/api/pith-number/A2QJHKSYJH2JFDTBIGC6D7SAXW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A2QJHKSYJH2JFDTBIGC6D7SAXW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A2QJHKSYJH2JFDTBIGC6D7SAXW/action/storage_attestation","attest_author":"https://pith.science/pith/A2QJHKSYJH2JFDTBIGC6D7SAXW/action/author_attestation","sign_citation":"https://pith.science/pith/A2QJHKSYJH2JFDTBIGC6D7SAXW/action/citation_signature","submit_replication":"https://pith.science/pith/A2QJHKSYJH2JFDTBIGC6D7SAXW/action/replication_record"}},"created_at":"2026-07-05T07:12:51.948202+00:00","updated_at":"2026-07-05T07:12:51.948202+00:00"}