{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:F4UBLMQ3SAF5I4FZVYG2BZY3PF","short_pith_number":"pith:F4UBLMQ3","schema_version":"1.0","canonical_sha256":"2f2815b21b900bd470b9ae0da0e71b79635cb8f8824d0cfd0c544993d46dda0c","source":{"kind":"arxiv","id":"2606.24000","version":1},"attestation_state":"computed","paper":{"title":"Cyclic Denoising Reveals Ultrastable Memories in Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.CR","cs.CV"],"primary_cat":"cs.LG","authors_text":"Rishabh Sharma, Stefano Martiniani","submitted_at":"2026-06-22T23:11:38Z","abstract_excerpt":"We introduce cyclic denoising -- repeated forward and reverse diffusion at controlled noise amplitudes -- as an extraction attack for image diffusion models. Inspired by random organization in disordered solids, cyclic denoising exposes regions of the learned distribution that are largely inaccessible to standard sampling. The dynamics drive samples toward attractors with a broad stability spectrum. The deepest attractors are ultrastable: they regenerate after near-total corruption and persist through thousands of noising-denoising cycles. Many of these attractors correspond to memorized train"},"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":"2606.24000","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-22T23:11:38Z","cross_cats_sorted":["cond-mat.dis-nn","cs.CR","cs.CV"],"title_canon_sha256":"7ce2503404d75e1bfdb7c805a89e757eac02afd2bc74431f7e56dcefd2b3ca2b","abstract_canon_sha256":"09934fb49fc9b3293a4f411d7b2c5c6b87b8a274128284cdf61390026fbdeceb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-24T00:14:32.687903Z","signature_b64":"SgX35W2A4JVmE7dsj0vT1DnjRtBwOEUDv9/u0npf8x9pSg+ea/dpvLUyKMncmW9MEP/ODOOTYqcAiJwFmKz9Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2f2815b21b900bd470b9ae0da0e71b79635cb8f8824d0cfd0c544993d46dda0c","last_reissued_at":"2026-06-24T00:14:32.687519Z","signature_status":"signed_v1","first_computed_at":"2026-06-24T00:14:32.687519Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cyclic Denoising Reveals Ultrastable Memories in Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.CR","cs.CV"],"primary_cat":"cs.LG","authors_text":"Rishabh Sharma, Stefano Martiniani","submitted_at":"2026-06-22T23:11:38Z","abstract_excerpt":"We introduce cyclic denoising -- repeated forward and reverse diffusion at controlled noise amplitudes -- as an extraction attack for image diffusion models. Inspired by random organization in disordered solids, cyclic denoising exposes regions of the learned distribution that are largely inaccessible to standard sampling. The dynamics drive samples toward attractors with a broad stability spectrum. The deepest attractors are ultrastable: they regenerate after near-total corruption and persist through thousands of noising-denoising cycles. Many of these attractors correspond to memorized train"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.24000","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/2606.24000/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":"2606.24000","created_at":"2026-06-24T00:14:32.687583+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.24000v1","created_at":"2026-06-24T00:14:32.687583+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.24000","created_at":"2026-06-24T00:14:32.687583+00:00"},{"alias_kind":"pith_short_12","alias_value":"F4UBLMQ3SAF5","created_at":"2026-06-24T00:14:32.687583+00:00"},{"alias_kind":"pith_short_16","alias_value":"F4UBLMQ3SAF5I4FZ","created_at":"2026-06-24T00:14:32.687583+00:00"},{"alias_kind":"pith_short_8","alias_value":"F4UBLMQ3","created_at":"2026-06-24T00:14:32.687583+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.29150","citing_title":"Flow Reasoning Models: Scaling Reasoning Through Iterative Self-Refinement","ref_index":28,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/F4UBLMQ3SAF5I4FZVYG2BZY3PF","json":"https://pith.science/pith/F4UBLMQ3SAF5I4FZVYG2BZY3PF.json","graph_json":"https://pith.science/api/pith-number/F4UBLMQ3SAF5I4FZVYG2BZY3PF/graph.json","events_json":"https://pith.science/api/pith-number/F4UBLMQ3SAF5I4FZVYG2BZY3PF/events.json","paper":"https://pith.science/paper/F4UBLMQ3"},"agent_actions":{"view_html":"https://pith.science/pith/F4UBLMQ3SAF5I4FZVYG2BZY3PF","download_json":"https://pith.science/pith/F4UBLMQ3SAF5I4FZVYG2BZY3PF.json","view_paper":"https://pith.science/paper/F4UBLMQ3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.24000&json=true","fetch_graph":"https://pith.science/api/pith-number/F4UBLMQ3SAF5I4FZVYG2BZY3PF/graph.json","fetch_events":"https://pith.science/api/pith-number/F4UBLMQ3SAF5I4FZVYG2BZY3PF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F4UBLMQ3SAF5I4FZVYG2BZY3PF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F4UBLMQ3SAF5I4FZVYG2BZY3PF/action/storage_attestation","attest_author":"https://pith.science/pith/F4UBLMQ3SAF5I4FZVYG2BZY3PF/action/author_attestation","sign_citation":"https://pith.science/pith/F4UBLMQ3SAF5I4FZVYG2BZY3PF/action/citation_signature","submit_replication":"https://pith.science/pith/F4UBLMQ3SAF5I4FZVYG2BZY3PF/action/replication_record"}},"created_at":"2026-06-24T00:14:32.687583+00:00","updated_at":"2026-06-24T00:14:32.687583+00:00"}