{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QWE7CP4JZX375DTIZNE5MWROJG","short_pith_number":"pith:QWE7CP4J","schema_version":"1.0","canonical_sha256":"8589f13f89cdf7fe8e68cb49d65a2e4991bc54d9b6f2dcdf5b5579c1a13e2c68","source":{"kind":"arxiv","id":"2605.18719","version":1},"attestation_state":"computed","paper":{"title":"SafeDiffusion-R1: Online Reward Steering for Safe Diffusion Post-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Abhishek Basu, Ankan Deria, Fahad Shamshad, Hisham Cholakkal, Karthik Nandakumar, Komal Kumar","submitted_at":"2026-05-18T17:50:04Z","abstract_excerpt":"Diffusion models have been widely studied for removing unsafe content learned during pre-training. Existing methods require expensive supervised data, either unsafe-text paired with safe-image groundtruth or negative/positive image pairs, making them impractical to scale. Furthermore, offline reinforcement learning and supervised fine-tuning approaches that generate synthetic data offline suffer from catastrophic forgetting, degrading generation quality. We propose a novel online reinforcement learning framework that addresses both data scarcity and model degradation through post-training with"},"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":"2605.18719","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-18T17:50:04Z","cross_cats_sorted":[],"title_canon_sha256":"688ccdd1eb5de41375c1154406b5d8f3e9b21e68b937c7dc772c25a65fa96d30","abstract_canon_sha256":"92be38605f9d6f4c30bd09e5a3efa4ac33d0a1399e56ee855e6509c8282ad8b6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:06:16.826136Z","signature_b64":"RlFLn/AnN5VLvPdQlFkW3mPWnJZp8vIdzXe2vXC/dm99HSdEui9eSJLkckVZw3CPTvJxRNTO5AJBDMyOuiAUCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8589f13f89cdf7fe8e68cb49d65a2e4991bc54d9b6f2dcdf5b5579c1a13e2c68","last_reissued_at":"2026-05-20T00:06:16.825284Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:06:16.825284Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SafeDiffusion-R1: Online Reward Steering for Safe Diffusion Post-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Abhishek Basu, Ankan Deria, Fahad Shamshad, Hisham Cholakkal, Karthik Nandakumar, Komal Kumar","submitted_at":"2026-05-18T17:50:04Z","abstract_excerpt":"Diffusion models have been widely studied for removing unsafe content learned during pre-training. Existing methods require expensive supervised data, either unsafe-text paired with safe-image groundtruth or negative/positive image pairs, making them impractical to scale. Furthermore, offline reinforcement learning and supervised fine-tuning approaches that generate synthetic data offline suffer from catastrophic forgetting, degrading generation quality. We propose a novel online reinforcement learning framework that addresses both data scarcity and model degradation through post-training with"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.18719","kind":"arxiv","version":1},"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/2605.18719/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-20T00:01:59.043508Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"629a37b62915818045a3708bd2df71c4857eca19b790a66ce153ae022e43cd85"},"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":"2605.18719","created_at":"2026-05-20T00:06:16.825454+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.18719v1","created_at":"2026-05-20T00:06:16.825454+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.18719","created_at":"2026-05-20T00:06:16.825454+00:00"},{"alias_kind":"pith_short_12","alias_value":"QWE7CP4JZX37","created_at":"2026-05-20T00:06:16.825454+00:00"},{"alias_kind":"pith_short_16","alias_value":"QWE7CP4JZX375DTI","created_at":"2026-05-20T00:06:16.825454+00:00"},{"alias_kind":"pith_short_8","alias_value":"QWE7CP4J","created_at":"2026-05-20T00:06:16.825454+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QWE7CP4JZX375DTIZNE5MWROJG","json":"https://pith.science/pith/QWE7CP4JZX375DTIZNE5MWROJG.json","graph_json":"https://pith.science/api/pith-number/QWE7CP4JZX375DTIZNE5MWROJG/graph.json","events_json":"https://pith.science/api/pith-number/QWE7CP4JZX375DTIZNE5MWROJG/events.json","paper":"https://pith.science/paper/QWE7CP4J"},"agent_actions":{"view_html":"https://pith.science/pith/QWE7CP4JZX375DTIZNE5MWROJG","download_json":"https://pith.science/pith/QWE7CP4JZX375DTIZNE5MWROJG.json","view_paper":"https://pith.science/paper/QWE7CP4J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.18719&json=true","fetch_graph":"https://pith.science/api/pith-number/QWE7CP4JZX375DTIZNE5MWROJG/graph.json","fetch_events":"https://pith.science/api/pith-number/QWE7CP4JZX375DTIZNE5MWROJG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QWE7CP4JZX375DTIZNE5MWROJG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QWE7CP4JZX375DTIZNE5MWROJG/action/storage_attestation","attest_author":"https://pith.science/pith/QWE7CP4JZX375DTIZNE5MWROJG/action/author_attestation","sign_citation":"https://pith.science/pith/QWE7CP4JZX375DTIZNE5MWROJG/action/citation_signature","submit_replication":"https://pith.science/pith/QWE7CP4JZX375DTIZNE5MWROJG/action/replication_record"}},"created_at":"2026-05-20T00:06:16.825454+00:00","updated_at":"2026-05-20T00:06:16.825454+00:00"}