{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ZXP2FNIOQZ37DIJQVRKXQYW7IQ","short_pith_number":"pith:ZXP2FNIO","schema_version":"1.0","canonical_sha256":"cddfa2b50e8677f1a130ac557862df443f56e3791cf0f9d7bd769e9ebc119d42","source":{"kind":"arxiv","id":"2504.13203","version":2},"attestation_state":"computed","paper":{"title":"X-Teaming: Multi-Turn Jailbreaks and Defenses with Adaptive Multi-Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG","cs.MA"],"primary_cat":"cs.CR","authors_text":"Genglin Liu, Hamid Palangi, James Shiffer, Kai-Wei Chang, Liwei Jiang, Md Rizwan Parvez, Saadia Gabriel, Salman Rahman, Sheriff Issaka, Yejin Choi","submitted_at":"2025-04-15T16:11:28Z","abstract_excerpt":"Multi-turn interactions with language models (LMs) pose critical safety risks, as harmful intent can be strategically spread across exchanges. Yet, the vast majority of prior work has focused on single-turn safety, while adaptability and diversity remain among the key challenges of multi-turn red-teaming. To address these challenges, we present X-Teaming, a scalable framework that systematically explores how seemingly harmless interactions escalate into harmful outcomes and generates corresponding attack scenarios. X-Teaming employs collaborative agents for planning, attack optimization, and v"},"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":"2504.13203","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2025-04-15T16:11:28Z","cross_cats_sorted":["cs.AI","cs.CL","cs.LG","cs.MA"],"title_canon_sha256":"e4f2f540fb2381a4ba8fb960721fa248a03f463c7dcde5aac1ff1daad5dea625","abstract_canon_sha256":"c1db50f0db0d5160021c592ceae183687aa75ed162f924532902235e82519ec6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:57:56.053839Z","signature_b64":"4cQZUbO+pnvc8kskp5eD++cqgNhgn0TLE+B+Egit58XKeISyH34UEGSaiFsSd5CcuKg+BEvXyuJAGCbsxq5QAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cddfa2b50e8677f1a130ac557862df443f56e3791cf0f9d7bd769e9ebc119d42","last_reissued_at":"2026-07-05T11:57:56.053367Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:57:56.053367Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"X-Teaming: Multi-Turn Jailbreaks and Defenses with Adaptive Multi-Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG","cs.MA"],"primary_cat":"cs.CR","authors_text":"Genglin Liu, Hamid Palangi, James Shiffer, Kai-Wei Chang, Liwei Jiang, Md Rizwan Parvez, Saadia Gabriel, Salman Rahman, Sheriff Issaka, Yejin Choi","submitted_at":"2025-04-15T16:11:28Z","abstract_excerpt":"Multi-turn interactions with language models (LMs) pose critical safety risks, as harmful intent can be strategically spread across exchanges. Yet, the vast majority of prior work has focused on single-turn safety, while adaptability and diversity remain among the key challenges of multi-turn red-teaming. To address these challenges, we present X-Teaming, a scalable framework that systematically explores how seemingly harmless interactions escalate into harmful outcomes and generates corresponding attack scenarios. X-Teaming employs collaborative agents for planning, attack optimization, and v"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.13203","kind":"arxiv","version":2},"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/2504.13203/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":"2504.13203","created_at":"2026-07-05T11:57:56.053422+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.13203v2","created_at":"2026-07-05T11:57:56.053422+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.13203","created_at":"2026-07-05T11:57:56.053422+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZXP2FNIOQZ37","created_at":"2026-07-05T11:57:56.053422+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZXP2FNIOQZ37DIJQ","created_at":"2026-07-05T11:57:56.053422+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZXP2FNIO","created_at":"2026-07-05T11:57:56.053422+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":14,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.10484","citing_title":"AgentCanary: A Security Evaluation Framework for Autonomous AI Agents in Real Executable Environments","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11470","citing_title":"The Periodic Table of LLM Reasoning: A Structured Survey of Reasoning Paradigms, Methods, and Failure Modes","ref_index":201,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31748","citing_title":"Addressing Over-Refusal in LLMs with Competing Rewards","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29224","citing_title":"Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2511.12710","citing_title":"Evolve the Method, Not the Prompts: Evolutionary Synthesis of Jailbreak Attacks on LLMs","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2508.21457","citing_title":"SoK: Exposing the Generation and Detection Gaps in LLM-Generated Phishing","ref_index":76,"is_internal_anchor":false},{"citing_arxiv_id":"2604.01473","citing_title":"SelfGrader: LLM Jailbreak Detection via Anchored Token-Level Logits","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05630","citing_title":"One Turn Too Late: Response-Aware Defense Against Hidden Malicious Intent in Multi-Turn Dialogue","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11002","citing_title":"MT-JailBench: A Modular Benchmark for Understanding Multi-Turn Jailbreak Attacks","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08277","citing_title":"Mitigating Many-shot Jailbreak Attacks with One Single Demonstration","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05630","citing_title":"One Turn Too Late: Response-Aware Defense Against Hidden Malicious Intent in Multi-Turn Dialogue","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01687","citing_title":"MultiBreak: A Scalable and Diverse Multi-turn Jailbreak Benchmark for Evaluating LLM Safety","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11309","citing_title":"The Salami Slicing Threat: Exploiting Cumulative Risks in LLM Systems","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16543","citing_title":"Conjunctive Prompt Attacks in Multi-Agent LLM Systems","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZXP2FNIOQZ37DIJQVRKXQYW7IQ","json":"https://pith.science/pith/ZXP2FNIOQZ37DIJQVRKXQYW7IQ.json","graph_json":"https://pith.science/api/pith-number/ZXP2FNIOQZ37DIJQVRKXQYW7IQ/graph.json","events_json":"https://pith.science/api/pith-number/ZXP2FNIOQZ37DIJQVRKXQYW7IQ/events.json","paper":"https://pith.science/paper/ZXP2FNIO"},"agent_actions":{"view_html":"https://pith.science/pith/ZXP2FNIOQZ37DIJQVRKXQYW7IQ","download_json":"https://pith.science/pith/ZXP2FNIOQZ37DIJQVRKXQYW7IQ.json","view_paper":"https://pith.science/paper/ZXP2FNIO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.13203&json=true","fetch_graph":"https://pith.science/api/pith-number/ZXP2FNIOQZ37DIJQVRKXQYW7IQ/graph.json","fetch_events":"https://pith.science/api/pith-number/ZXP2FNIOQZ37DIJQVRKXQYW7IQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZXP2FNIOQZ37DIJQVRKXQYW7IQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZXP2FNIOQZ37DIJQVRKXQYW7IQ/action/storage_attestation","attest_author":"https://pith.science/pith/ZXP2FNIOQZ37DIJQVRKXQYW7IQ/action/author_attestation","sign_citation":"https://pith.science/pith/ZXP2FNIOQZ37DIJQVRKXQYW7IQ/action/citation_signature","submit_replication":"https://pith.science/pith/ZXP2FNIOQZ37DIJQVRKXQYW7IQ/action/replication_record"}},"created_at":"2026-07-05T11:57:56.053422+00:00","updated_at":"2026-07-05T11:57:56.053422+00:00"}