{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NV5RCK756IO6NIV6UWWLEAOSYO","short_pith_number":"pith:NV5RCK75","schema_version":"1.0","canonical_sha256":"6d7b112bfdf21de6a2bea5acb201d2c3bfcf2705ab61d04f8000349a6d2ecb1f","source":{"kind":"arxiv","id":"2606.03136","version":1},"attestation_state":"computed","paper":{"title":"PsychoPass: Geometric Profiling of Multi-Turn Adversarial LLM Conversations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Muberra Ozmen, Subhabrata Majumdar","submitted_at":"2026-06-02T04:24:20Z","abstract_excerpt":"Multi-turn jailbreak attacks on large language models (LLMs) reveal a mismatch in current guardrails: they operate on individual turns, while attacks unfold as trajectories across conversations. We propose a shift from content to dynamics, modeling conversations as paths in representation space and asking whether adversarial intent is encoded early in their geometry. We introduce PsychoPass, a framework that extracts geometric features from conversation trajectories in embedding space to predict a potential attack before harmful content is produced. These features achieve near-perfect performa"},"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":"2606.03136","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2026-06-02T04:24:20Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"6ccd0292ea71a8fe7f2444121f797f214b71096cbba79ce988daeb41ab97cc9b","abstract_canon_sha256":"a06ceaf5782f78553111934235c9f7b9e25ecb30ac053c2077f1f04e82bccde5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-03T01:05:32.687045Z","signature_b64":"Yhb6B1rE422Vowl5PcAuo0tOGKAXwN8DC1/IaXhRAvLvyQwlQUyLcGeKCxjsX55iDg4gJg8DFGfx1qxgT0GWDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6d7b112bfdf21de6a2bea5acb201d2c3bfcf2705ab61d04f8000349a6d2ecb1f","last_reissued_at":"2026-06-03T01:05:32.686636Z","signature_status":"signed_v1","first_computed_at":"2026-06-03T01:05:32.686636Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PsychoPass: Geometric Profiling of Multi-Turn Adversarial LLM Conversations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Muberra Ozmen, Subhabrata Majumdar","submitted_at":"2026-06-02T04:24:20Z","abstract_excerpt":"Multi-turn jailbreak attacks on large language models (LLMs) reveal a mismatch in current guardrails: they operate on individual turns, while attacks unfold as trajectories across conversations. We propose a shift from content to dynamics, modeling conversations as paths in representation space and asking whether adversarial intent is encoded early in their geometry. We introduce PsychoPass, a framework that extracts geometric features from conversation trajectories in embedding space to predict a potential attack before harmful content is produced. These features achieve near-perfect performa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.03136","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/2606.03136/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":"2606.03136","created_at":"2026-06-03T01:05:32.686694+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.03136v1","created_at":"2026-06-03T01:05:32.686694+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.03136","created_at":"2026-06-03T01:05:32.686694+00:00"},{"alias_kind":"pith_short_12","alias_value":"NV5RCK756IO6","created_at":"2026-06-03T01:05:32.686694+00:00"},{"alias_kind":"pith_short_16","alias_value":"NV5RCK756IO6NIV6","created_at":"2026-06-03T01:05:32.686694+00:00"},{"alias_kind":"pith_short_8","alias_value":"NV5RCK75","created_at":"2026-06-03T01:05:32.686694+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/NV5RCK756IO6NIV6UWWLEAOSYO","json":"https://pith.science/pith/NV5RCK756IO6NIV6UWWLEAOSYO.json","graph_json":"https://pith.science/api/pith-number/NV5RCK756IO6NIV6UWWLEAOSYO/graph.json","events_json":"https://pith.science/api/pith-number/NV5RCK756IO6NIV6UWWLEAOSYO/events.json","paper":"https://pith.science/paper/NV5RCK75"},"agent_actions":{"view_html":"https://pith.science/pith/NV5RCK756IO6NIV6UWWLEAOSYO","download_json":"https://pith.science/pith/NV5RCK756IO6NIV6UWWLEAOSYO.json","view_paper":"https://pith.science/paper/NV5RCK75","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.03136&json=true","fetch_graph":"https://pith.science/api/pith-number/NV5RCK756IO6NIV6UWWLEAOSYO/graph.json","fetch_events":"https://pith.science/api/pith-number/NV5RCK756IO6NIV6UWWLEAOSYO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NV5RCK756IO6NIV6UWWLEAOSYO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NV5RCK756IO6NIV6UWWLEAOSYO/action/storage_attestation","attest_author":"https://pith.science/pith/NV5RCK756IO6NIV6UWWLEAOSYO/action/author_attestation","sign_citation":"https://pith.science/pith/NV5RCK756IO6NIV6UWWLEAOSYO/action/citation_signature","submit_replication":"https://pith.science/pith/NV5RCK756IO6NIV6UWWLEAOSYO/action/replication_record"}},"created_at":"2026-06-03T01:05:32.686694+00:00","updated_at":"2026-06-03T01:05:32.686694+00:00"}