{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:B6QOVNIT7RNCUIPUL2KXKUTCNY","short_pith_number":"pith:B6QOVNIT","schema_version":"1.0","canonical_sha256":"0fa0eab513fc5a2a21f45e957552626e0822dd4a9adafdc2848c74cf9dfbd72c","source":{"kind":"arxiv","id":"2410.06025","version":3},"attestation_state":"computed","paper":{"title":"Shielded Diffusion: Generating Novel and Diverse Images using Sparse Repellency","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"cs.CV","authors_text":"Eugene Ndiaye, James Thornton, Louis B\\'ethune, Marco Cuturi, Michael Kirchhof, Pierre Ablin","submitted_at":"2024-10-08T13:26:32Z","abstract_excerpt":"The adoption of text-to-image diffusion models raises concerns over reliability, drawing scrutiny under the lens of various metrics like calibration, fairness, or compute efficiency. We focus in this work on two issues that arise when deploying these models: a lack of diversity when prompting images, and a tendency to recreate images from the training set. To solve both problems, we propose a method that coaxes the sampled trajectories of pretrained diffusion models to land on images that fall outside of a reference set. We achieve this by adding repellency terms to the diffusion SDE throughou"},"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":"2410.06025","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-10-08T13:26:32Z","cross_cats_sorted":["cs.LG","stat.ML"],"title_canon_sha256":"37a6ebfe1c2c1ad1df9f76757d0659f515ba5dca95f0d953aa1205b6965de803","abstract_canon_sha256":"25b106c6c820db30a4e05123678a7ed0a9319920e7d92fa9c0e16b5a529d54c3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:11:20.851082Z","signature_b64":"ZNCmaSL79LxsAH3kCWi6QJjauAhlKV1waw1/tVYu1cQ4q74xjsfy/1MYnBoq1T9UJQc6PL4XTlRGfPbY5YD2BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0fa0eab513fc5a2a21f45e957552626e0822dd4a9adafdc2848c74cf9dfbd72c","last_reissued_at":"2026-07-05T11:11:20.850578Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:11:20.850578Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Shielded Diffusion: Generating Novel and Diverse Images using Sparse Repellency","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"cs.CV","authors_text":"Eugene Ndiaye, James Thornton, Louis B\\'ethune, Marco Cuturi, Michael Kirchhof, Pierre Ablin","submitted_at":"2024-10-08T13:26:32Z","abstract_excerpt":"The adoption of text-to-image diffusion models raises concerns over reliability, drawing scrutiny under the lens of various metrics like calibration, fairness, or compute efficiency. We focus in this work on two issues that arise when deploying these models: a lack of diversity when prompting images, and a tendency to recreate images from the training set. To solve both problems, we propose a method that coaxes the sampled trajectories of pretrained diffusion models to land on images that fall outside of a reference set. We achieve this by adding repellency terms to the diffusion SDE throughou"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.06025","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/2410.06025/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":"2410.06025","created_at":"2026-07-05T11:11:20.850637+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.06025v3","created_at":"2026-07-05T11:11:20.850637+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.06025","created_at":"2026-07-05T11:11:20.850637+00:00"},{"alias_kind":"pith_short_12","alias_value":"B6QOVNIT7RNC","created_at":"2026-07-05T11:11:20.850637+00:00"},{"alias_kind":"pith_short_16","alias_value":"B6QOVNIT7RNCUIPU","created_at":"2026-07-05T11:11:20.850637+00:00"},{"alias_kind":"pith_short_8","alias_value":"B6QOVNIT","created_at":"2026-07-05T11:11:20.850637+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27371","citing_title":"Don't Settle at the Mode! Mitigating Diversity Collapse in Pretrained Flow Models via Feature Self-Guidance","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23267","citing_title":"Safe Few-Step Generation via Velocity Editing","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06813","citing_title":"Breaking the Lock-in: Diversifying Text-to-Image Generation via Representation Modulation","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2601.00090","citing_title":"It's Never Too Late: Noise Optimization for Collapse Recovery in Trained Diffusion Models","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11494","citing_title":"STRIDE: Training-Free Diversity Guidance via PCA-Directed Feature Perturbation in Single-Step Diffusion Models","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08116","citing_title":"The Safety-Aware Denoiser for Text Diffusion Models","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/B6QOVNIT7RNCUIPUL2KXKUTCNY","json":"https://pith.science/pith/B6QOVNIT7RNCUIPUL2KXKUTCNY.json","graph_json":"https://pith.science/api/pith-number/B6QOVNIT7RNCUIPUL2KXKUTCNY/graph.json","events_json":"https://pith.science/api/pith-number/B6QOVNIT7RNCUIPUL2KXKUTCNY/events.json","paper":"https://pith.science/paper/B6QOVNIT"},"agent_actions":{"view_html":"https://pith.science/pith/B6QOVNIT7RNCUIPUL2KXKUTCNY","download_json":"https://pith.science/pith/B6QOVNIT7RNCUIPUL2KXKUTCNY.json","view_paper":"https://pith.science/paper/B6QOVNIT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.06025&json=true","fetch_graph":"https://pith.science/api/pith-number/B6QOVNIT7RNCUIPUL2KXKUTCNY/graph.json","fetch_events":"https://pith.science/api/pith-number/B6QOVNIT7RNCUIPUL2KXKUTCNY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B6QOVNIT7RNCUIPUL2KXKUTCNY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B6QOVNIT7RNCUIPUL2KXKUTCNY/action/storage_attestation","attest_author":"https://pith.science/pith/B6QOVNIT7RNCUIPUL2KXKUTCNY/action/author_attestation","sign_citation":"https://pith.science/pith/B6QOVNIT7RNCUIPUL2KXKUTCNY/action/citation_signature","submit_replication":"https://pith.science/pith/B6QOVNIT7RNCUIPUL2KXKUTCNY/action/replication_record"}},"created_at":"2026-07-05T11:11:20.850637+00:00","updated_at":"2026-07-05T11:11:20.850637+00:00"}