{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ETKSKLREEG6QIOY25PK2XPVIPR","short_pith_number":"pith:ETKSKLRE","schema_version":"1.0","canonical_sha256":"24d5252e2421bd043b1aebd5abbea87c734de3421586e6c013252d68809f5004","source":{"kind":"arxiv","id":"2504.01278","version":1},"attestation_state":"computed","paper":{"title":"Strategize Globally, Adapt Locally: A Multi-Turn Red Teaming Agent with Dual-Level Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Ninareh Mehrabi, Rahul Gupta, Ruoxi Jia, Si Chen, Xiao Yu, Zhou Yu","submitted_at":"2025-04-02T01:06:19Z","abstract_excerpt":"The exploitation of large language models (LLMs) for malicious purposes poses significant security risks as these models become more powerful and widespread. While most existing red-teaming frameworks focus on single-turn attacks, real-world adversaries typically operate in multi-turn scenarios, iteratively probing for vulnerabilities and adapting their prompts based on threat model responses. In this paper, we propose \\AlgName, a novel multi-turn red-teaming agent that emulates sophisticated human attackers through complementary learning dimensions: global tactic-wise learning that accumulate"},"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":"2504.01278","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-04-02T01:06:19Z","cross_cats_sorted":[],"title_canon_sha256":"bfc16e0c87171a3c83007f75ecceff33af218f81ab710481722450c9d3880ead","abstract_canon_sha256":"58702eec8c05a4d92205c1c73d90836b86657f096f39923ad99577a2edea03ee"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:43:12.055575Z","signature_b64":"KiabeN3JBCUExWF5btNhwv9VLXaQcK6oHxAn3SEfxpEgd+ee3grXmAFMQR6UuR5DUUWJico0LDJXs8CTzVZUBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"24d5252e2421bd043b1aebd5abbea87c734de3421586e6c013252d68809f5004","last_reissued_at":"2026-07-05T10:43:12.055050Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:43:12.055050Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Strategize Globally, Adapt Locally: A Multi-Turn Red Teaming Agent with Dual-Level Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Ninareh Mehrabi, Rahul Gupta, Ruoxi Jia, Si Chen, Xiao Yu, Zhou Yu","submitted_at":"2025-04-02T01:06:19Z","abstract_excerpt":"The exploitation of large language models (LLMs) for malicious purposes poses significant security risks as these models become more powerful and widespread. While most existing red-teaming frameworks focus on single-turn attacks, real-world adversaries typically operate in multi-turn scenarios, iteratively probing for vulnerabilities and adapting their prompts based on threat model responses. In this paper, we propose \\AlgName, a novel multi-turn red-teaming agent that emulates sophisticated human attackers through complementary learning dimensions: global tactic-wise learning that accumulate"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.01278","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/2504.01278/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":"2504.01278","created_at":"2026-07-05T10:43:12.055099+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.01278v1","created_at":"2026-07-05T10:43:12.055099+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.01278","created_at":"2026-07-05T10:43:12.055099+00:00"},{"alias_kind":"pith_short_12","alias_value":"ETKSKLREEG6Q","created_at":"2026-07-05T10:43:12.055099+00:00"},{"alias_kind":"pith_short_16","alias_value":"ETKSKLREEG6QIOY2","created_at":"2026-07-05T10:43:12.055099+00:00"},{"alias_kind":"pith_short_8","alias_value":"ETKSKLRE","created_at":"2026-07-05T10:43:12.055099+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.01117","citing_title":"SoK: Intent-Oriented Systematization of Multi-Turn LLM Jailbreaks","ref_index":59,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ETKSKLREEG6QIOY25PK2XPVIPR","json":"https://pith.science/pith/ETKSKLREEG6QIOY25PK2XPVIPR.json","graph_json":"https://pith.science/api/pith-number/ETKSKLREEG6QIOY25PK2XPVIPR/graph.json","events_json":"https://pith.science/api/pith-number/ETKSKLREEG6QIOY25PK2XPVIPR/events.json","paper":"https://pith.science/paper/ETKSKLRE"},"agent_actions":{"view_html":"https://pith.science/pith/ETKSKLREEG6QIOY25PK2XPVIPR","download_json":"https://pith.science/pith/ETKSKLREEG6QIOY25PK2XPVIPR.json","view_paper":"https://pith.science/paper/ETKSKLRE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.01278&json=true","fetch_graph":"https://pith.science/api/pith-number/ETKSKLREEG6QIOY25PK2XPVIPR/graph.json","fetch_events":"https://pith.science/api/pith-number/ETKSKLREEG6QIOY25PK2XPVIPR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ETKSKLREEG6QIOY25PK2XPVIPR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ETKSKLREEG6QIOY25PK2XPVIPR/action/storage_attestation","attest_author":"https://pith.science/pith/ETKSKLREEG6QIOY25PK2XPVIPR/action/author_attestation","sign_citation":"https://pith.science/pith/ETKSKLREEG6QIOY25PK2XPVIPR/action/citation_signature","submit_replication":"https://pith.science/pith/ETKSKLREEG6QIOY25PK2XPVIPR/action/replication_record"}},"created_at":"2026-07-05T10:43:12.055099+00:00","updated_at":"2026-07-05T10:43:12.055099+00:00"}