{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:467VBC5WZ23L4ZLS7PGXMR2XMU","short_pith_number":"pith:467VBC5W","schema_version":"1.0","canonical_sha256":"e7bf508bb6ceb6be6572fbcd76475765315536d79fa3e92864ae307124b0d2a9","source":{"kind":"arxiv","id":"2505.06284","version":1},"attestation_state":"computed","paper":{"title":"DMRL: Data- and Model-aware Reward Learning for Data Extraction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Ruoxi Cheng, Zhiqiang Wang","submitted_at":"2025-05-07T07:21:37Z","abstract_excerpt":"Large language models (LLMs) are inherently vulnerable to unintended privacy breaches. Consequently, systematic red-teaming research is essential for developing robust defense mechanisms. However, current data extraction methods suffer from several limitations: (1) rely on dataset duplicates (addressable via deduplication), (2) depend on prompt engineering (now countered by detection and defense), and (3) rely on random-search adversarial generation. To address these challenges, we propose DMRL, a Data- and Model-aware Reward Learning approach for data extraction. This technique leverages inve"},"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":"2505.06284","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-05-07T07:21:37Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"6297b2def7b58dc41c0e959d4de6124ce660d0d18c7f1c6423d672dee0a4dcf6","abstract_canon_sha256":"5af34457e41bc715348a91857cf52e3f282c682fed778c151a730203743016c7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:01:06.614185Z","signature_b64":"RuYimNc+yDYkbeh2N6ZxsoGRj5F6TgcxTcPIqxGEfQkz0tbImWWEiMI3NqAqvhur879PXLksCWXjLFWKFXKfCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e7bf508bb6ceb6be6572fbcd76475765315536d79fa3e92864ae307124b0d2a9","last_reissued_at":"2026-07-05T11:01:06.613700Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:01:06.613700Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DMRL: Data- and Model-aware Reward Learning for Data Extraction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Ruoxi Cheng, Zhiqiang Wang","submitted_at":"2025-05-07T07:21:37Z","abstract_excerpt":"Large language models (LLMs) are inherently vulnerable to unintended privacy breaches. Consequently, systematic red-teaming research is essential for developing robust defense mechanisms. However, current data extraction methods suffer from several limitations: (1) rely on dataset duplicates (addressable via deduplication), (2) depend on prompt engineering (now countered by detection and defense), and (3) rely on random-search adversarial generation. To address these challenges, we propose DMRL, a Data- and Model-aware Reward Learning approach for data extraction. This technique leverages inve"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.06284","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/2505.06284/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":"2505.06284","created_at":"2026-07-05T11:01:06.613759+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.06284v1","created_at":"2026-07-05T11:01:06.613759+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.06284","created_at":"2026-07-05T11:01:06.613759+00:00"},{"alias_kind":"pith_short_12","alias_value":"467VBC5WZ23L","created_at":"2026-07-05T11:01:06.613759+00:00"},{"alias_kind":"pith_short_16","alias_value":"467VBC5WZ23L4ZLS","created_at":"2026-07-05T11:01:06.613759+00:00"},{"alias_kind":"pith_short_8","alias_value":"467VBC5W","created_at":"2026-07-05T11:01:06.613759+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/467VBC5WZ23L4ZLS7PGXMR2XMU","json":"https://pith.science/pith/467VBC5WZ23L4ZLS7PGXMR2XMU.json","graph_json":"https://pith.science/api/pith-number/467VBC5WZ23L4ZLS7PGXMR2XMU/graph.json","events_json":"https://pith.science/api/pith-number/467VBC5WZ23L4ZLS7PGXMR2XMU/events.json","paper":"https://pith.science/paper/467VBC5W"},"agent_actions":{"view_html":"https://pith.science/pith/467VBC5WZ23L4ZLS7PGXMR2XMU","download_json":"https://pith.science/pith/467VBC5WZ23L4ZLS7PGXMR2XMU.json","view_paper":"https://pith.science/paper/467VBC5W","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.06284&json=true","fetch_graph":"https://pith.science/api/pith-number/467VBC5WZ23L4ZLS7PGXMR2XMU/graph.json","fetch_events":"https://pith.science/api/pith-number/467VBC5WZ23L4ZLS7PGXMR2XMU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/467VBC5WZ23L4ZLS7PGXMR2XMU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/467VBC5WZ23L4ZLS7PGXMR2XMU/action/storage_attestation","attest_author":"https://pith.science/pith/467VBC5WZ23L4ZLS7PGXMR2XMU/action/author_attestation","sign_citation":"https://pith.science/pith/467VBC5WZ23L4ZLS7PGXMR2XMU/action/citation_signature","submit_replication":"https://pith.science/pith/467VBC5WZ23L4ZLS7PGXMR2XMU/action/replication_record"}},"created_at":"2026-07-05T11:01:06.613759+00:00","updated_at":"2026-07-05T11:01:06.613759+00:00"}