{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:6IRIUQTTBUEVAK4BY2OZQJPOGF","short_pith_number":"pith:6IRIUQTT","schema_version":"1.0","canonical_sha256":"f2228a42730d09502b81c69d9825ee314566ed9da4673111d3c114572bc1baa3","source":{"kind":"arxiv","id":"2605.27671","version":1},"attestation_state":"computed","paper":{"title":"Evolving and Detecting Multi-Turn Deception using Geometric Signatures","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Mary L. Cummings, Surender Suresh Kumar","submitted_at":"2026-05-26T20:48:33Z","abstract_excerpt":"Safety defenses for large language models (LLMs) are typically trained and evaluated on single-turn prompts, yet real attacks often unfold as indirect, multi-turn probing. To defend against this more nuanced form of deception, we present a unified pipeline that generates realistic multi-turn deceptive question sets via multi-objective genetic prompt optimization with co-evolving mutation operators. We validate this dataset through a human study, which also revealed that early generations yielded the most convincing deception and practical constraints such as adherence filtering and ordering ef"},"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":"2605.27671","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-26T20:48:33Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"2f2ed79021f2ff59cac2754032eae96c5f9415a60bcb2c30dfb1e780a041df3b","abstract_canon_sha256":"75a695e9903df76fe40d842fd205ec72818c5e7cc1d051065f306a5113070b2b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:45.820597Z","signature_b64":"wOKerQqaJZqFQMXs7aTiBlvcFDFPdEx84m6GykDX+Fr/6MEWdAI2a6ZSSwgmyTFxouhM/lQos3kpgip2raCdAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f2228a42730d09502b81c69d9825ee314566ed9da4673111d3c114572bc1baa3","last_reissued_at":"2026-05-28T01:04:45.820200Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:45.820200Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Evolving and Detecting Multi-Turn Deception using Geometric Signatures","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Mary L. Cummings, Surender Suresh Kumar","submitted_at":"2026-05-26T20:48:33Z","abstract_excerpt":"Safety defenses for large language models (LLMs) are typically trained and evaluated on single-turn prompts, yet real attacks often unfold as indirect, multi-turn probing. To defend against this more nuanced form of deception, we present a unified pipeline that generates realistic multi-turn deceptive question sets via multi-objective genetic prompt optimization with co-evolving mutation operators. We validate this dataset through a human study, which also revealed that early generations yielded the most convincing deception and practical constraints such as adherence filtering and ordering ef"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.27671","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/2605.27671/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":"2605.27671","created_at":"2026-05-28T01:04:45.820255+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.27671v1","created_at":"2026-05-28T01:04:45.820255+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.27671","created_at":"2026-05-28T01:04:45.820255+00:00"},{"alias_kind":"pith_short_12","alias_value":"6IRIUQTTBUEV","created_at":"2026-05-28T01:04:45.820255+00:00"},{"alias_kind":"pith_short_16","alias_value":"6IRIUQTTBUEVAK4B","created_at":"2026-05-28T01:04:45.820255+00:00"},{"alias_kind":"pith_short_8","alias_value":"6IRIUQTT","created_at":"2026-05-28T01:04:45.820255+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/6IRIUQTTBUEVAK4BY2OZQJPOGF","json":"https://pith.science/pith/6IRIUQTTBUEVAK4BY2OZQJPOGF.json","graph_json":"https://pith.science/api/pith-number/6IRIUQTTBUEVAK4BY2OZQJPOGF/graph.json","events_json":"https://pith.science/api/pith-number/6IRIUQTTBUEVAK4BY2OZQJPOGF/events.json","paper":"https://pith.science/paper/6IRIUQTT"},"agent_actions":{"view_html":"https://pith.science/pith/6IRIUQTTBUEVAK4BY2OZQJPOGF","download_json":"https://pith.science/pith/6IRIUQTTBUEVAK4BY2OZQJPOGF.json","view_paper":"https://pith.science/paper/6IRIUQTT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.27671&json=true","fetch_graph":"https://pith.science/api/pith-number/6IRIUQTTBUEVAK4BY2OZQJPOGF/graph.json","fetch_events":"https://pith.science/api/pith-number/6IRIUQTTBUEVAK4BY2OZQJPOGF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6IRIUQTTBUEVAK4BY2OZQJPOGF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6IRIUQTTBUEVAK4BY2OZQJPOGF/action/storage_attestation","attest_author":"https://pith.science/pith/6IRIUQTTBUEVAK4BY2OZQJPOGF/action/author_attestation","sign_citation":"https://pith.science/pith/6IRIUQTTBUEVAK4BY2OZQJPOGF/action/citation_signature","submit_replication":"https://pith.science/pith/6IRIUQTTBUEVAK4BY2OZQJPOGF/action/replication_record"}},"created_at":"2026-05-28T01:04:45.820255+00:00","updated_at":"2026-05-28T01:04:45.820255+00:00"}