{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ZMEMRFXYCQAQTMV4LDYVC2Q3QL","short_pith_number":"pith:ZMEMRFXY","schema_version":"1.0","canonical_sha256":"cb08c896f8140109b2bc58f1516a1b82c9ce8477fef5acc6665476b6a0eb34f0","source":{"kind":"arxiv","id":"2406.19395","version":1},"attestation_state":"computed","paper":{"title":"Dataset Size Recovery from LoRA Weights","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Eliahu Horwitz, Jonathan Kahana, Mohammad Salama, Yedid Hoshen","submitted_at":"2024-06-27T17:59:53Z","abstract_excerpt":"Model inversion and membership inference attacks aim to reconstruct and verify the data which a model was trained on. However, they are not guaranteed to find all training samples as they do not know the size of the training set. In this paper, we introduce a new task: dataset size recovery, that aims to determine the number of samples used to train a model, directly from its weights. We then propose DSiRe, a method for recovering the number of images used to fine-tune a model, in the common case where fine-tuning uses LoRA. We discover that both the norm and the spectrum of the LoRA matrices "},"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":"2406.19395","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-06-27T17:59:53Z","cross_cats_sorted":[],"title_canon_sha256":"6fb7497e04dace821c73d4e8143f22809fad1286422ba73ff2c90cd67e91efba","abstract_canon_sha256":"b3734675dd6bcefbdc4a1dfaebee48430a042a271a4348717e05524fafce19d8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:37:33.901004Z","signature_b64":"rf7IfP0N0d326RaNTmbxBKE19uAiRYBHp/IODu95rl94V6gFbTyEDr30DReAvds7uHvL//oQwOpGsENT5kFRCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cb08c896f8140109b2bc58f1516a1b82c9ce8477fef5acc6665476b6a0eb34f0","last_reissued_at":"2026-07-05T08:37:33.900554Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:37:33.900554Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dataset Size Recovery from LoRA Weights","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Eliahu Horwitz, Jonathan Kahana, Mohammad Salama, Yedid Hoshen","submitted_at":"2024-06-27T17:59:53Z","abstract_excerpt":"Model inversion and membership inference attacks aim to reconstruct and verify the data which a model was trained on. However, they are not guaranteed to find all training samples as they do not know the size of the training set. In this paper, we introduce a new task: dataset size recovery, that aims to determine the number of samples used to train a model, directly from its weights. We then propose DSiRe, a method for recovering the number of images used to fine-tune a model, in the common case where fine-tuning uses LoRA. We discover that both the norm and the spectrum of the LoRA matrices "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.19395","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/2406.19395/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":"2406.19395","created_at":"2026-07-05T08:37:33.900607+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.19395v1","created_at":"2026-07-05T08:37:33.900607+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.19395","created_at":"2026-07-05T08:37:33.900607+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZMEMRFXYCQAQ","created_at":"2026-07-05T08:37:33.900607+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZMEMRFXYCQAQTMV4","created_at":"2026-07-05T08:37:33.900607+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZMEMRFXY","created_at":"2026-07-05T08:37:33.900607+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2411.14430","citing_title":"Stable Flow: Vital Layers for Training-Free Image Editing","ref_index":76,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZMEMRFXYCQAQTMV4LDYVC2Q3QL","json":"https://pith.science/pith/ZMEMRFXYCQAQTMV4LDYVC2Q3QL.json","graph_json":"https://pith.science/api/pith-number/ZMEMRFXYCQAQTMV4LDYVC2Q3QL/graph.json","events_json":"https://pith.science/api/pith-number/ZMEMRFXYCQAQTMV4LDYVC2Q3QL/events.json","paper":"https://pith.science/paper/ZMEMRFXY"},"agent_actions":{"view_html":"https://pith.science/pith/ZMEMRFXYCQAQTMV4LDYVC2Q3QL","download_json":"https://pith.science/pith/ZMEMRFXYCQAQTMV4LDYVC2Q3QL.json","view_paper":"https://pith.science/paper/ZMEMRFXY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.19395&json=true","fetch_graph":"https://pith.science/api/pith-number/ZMEMRFXYCQAQTMV4LDYVC2Q3QL/graph.json","fetch_events":"https://pith.science/api/pith-number/ZMEMRFXYCQAQTMV4LDYVC2Q3QL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZMEMRFXYCQAQTMV4LDYVC2Q3QL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZMEMRFXYCQAQTMV4LDYVC2Q3QL/action/storage_attestation","attest_author":"https://pith.science/pith/ZMEMRFXYCQAQTMV4LDYVC2Q3QL/action/author_attestation","sign_citation":"https://pith.science/pith/ZMEMRFXYCQAQTMV4LDYVC2Q3QL/action/citation_signature","submit_replication":"https://pith.science/pith/ZMEMRFXYCQAQTMV4LDYVC2Q3QL/action/replication_record"}},"created_at":"2026-07-05T08:37:33.900607+00:00","updated_at":"2026-07-05T08:37:33.900607+00:00"}