{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IUJUTK3PXTV5TZBR5RAW432HST","short_pith_number":"pith:IUJUTK3P","schema_version":"1.0","canonical_sha256":"451349ab6fbcebd9e431ec416e6f4794d8e418c03390b96cec1a7e303c17dd7d","source":{"kind":"arxiv","id":"2402.13148","version":3},"attestation_state":"computed","paper":{"title":"Defending Jailbreak Prompts via In-Context Adversarial Game","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Haomin Zhuang, Hongyan Bao, Kehan Guo, Xiangliang Zhang, Yufei Han, Yujun Zhou, Zhenwen Liang","submitted_at":"2024-02-20T17:04:06Z","abstract_excerpt":"Large Language Models (LLMs) demonstrate remarkable capabilities across diverse applications. However, concerns regarding their security, particularly the vulnerability to jailbreak attacks, persist. Drawing inspiration from adversarial training in deep learning and LLM agent learning processes, we introduce the In-Context Adversarial Game (ICAG) for defending against jailbreaks without the need for fine-tuning. ICAG leverages agent learning to conduct an adversarial game, aiming to dynamically extend knowledge to defend against jailbreaks. Unlike traditional methods that rely on static datase"},"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":"2402.13148","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-02-20T17:04:06Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"b80bbe63c9c6980b3f27fbc3f4a3c0f9361168debf58982837f3223e53e16c93","abstract_canon_sha256":"0fac502ddcf374f887f78c994b03d1d47e97241d06d34c8488ec765ce3450433"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:07.806253Z","signature_b64":"UXGBBTPSYj1d1PvjC2fDB4dFn+TyjAHu1LVMvldjK7K4eNOy+4e6KKApSugdrmpJTgdRViZ3O+zalT2u6kjTAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"451349ab6fbcebd9e431ec416e6f4794d8e418c03390b96cec1a7e303c17dd7d","last_reissued_at":"2026-07-05T10:18:07.805826Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:07.805826Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Defending Jailbreak Prompts via In-Context Adversarial Game","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Haomin Zhuang, Hongyan Bao, Kehan Guo, Xiangliang Zhang, Yufei Han, Yujun Zhou, Zhenwen Liang","submitted_at":"2024-02-20T17:04:06Z","abstract_excerpt":"Large Language Models (LLMs) demonstrate remarkable capabilities across diverse applications. However, concerns regarding their security, particularly the vulnerability to jailbreak attacks, persist. Drawing inspiration from adversarial training in deep learning and LLM agent learning processes, we introduce the In-Context Adversarial Game (ICAG) for defending against jailbreaks without the need for fine-tuning. ICAG leverages agent learning to conduct an adversarial game, aiming to dynamically extend knowledge to defend against jailbreaks. Unlike traditional methods that rely on static datase"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.13148","kind":"arxiv","version":3},"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/2402.13148/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":"2402.13148","created_at":"2026-07-05T10:18:07.805877+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.13148v3","created_at":"2026-07-05T10:18:07.805877+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.13148","created_at":"2026-07-05T10:18:07.805877+00:00"},{"alias_kind":"pith_short_12","alias_value":"IUJUTK3PXTV5","created_at":"2026-07-05T10:18:07.805877+00:00"},{"alias_kind":"pith_short_16","alias_value":"IUJUTK3PXTV5TZBR","created_at":"2026-07-05T10:18:07.805877+00:00"},{"alias_kind":"pith_short_8","alias_value":"IUJUTK3P","created_at":"2026-07-05T10:18:07.805877+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.08846","citing_title":"Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs","ref_index":132,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18493","citing_title":"Too Correct to Learn: Reinforcement Learning on Saturated Reasoning Data","ref_index":49,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IUJUTK3PXTV5TZBR5RAW432HST","json":"https://pith.science/pith/IUJUTK3PXTV5TZBR5RAW432HST.json","graph_json":"https://pith.science/api/pith-number/IUJUTK3PXTV5TZBR5RAW432HST/graph.json","events_json":"https://pith.science/api/pith-number/IUJUTK3PXTV5TZBR5RAW432HST/events.json","paper":"https://pith.science/paper/IUJUTK3P"},"agent_actions":{"view_html":"https://pith.science/pith/IUJUTK3PXTV5TZBR5RAW432HST","download_json":"https://pith.science/pith/IUJUTK3PXTV5TZBR5RAW432HST.json","view_paper":"https://pith.science/paper/IUJUTK3P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.13148&json=true","fetch_graph":"https://pith.science/api/pith-number/IUJUTK3PXTV5TZBR5RAW432HST/graph.json","fetch_events":"https://pith.science/api/pith-number/IUJUTK3PXTV5TZBR5RAW432HST/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IUJUTK3PXTV5TZBR5RAW432HST/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IUJUTK3PXTV5TZBR5RAW432HST/action/storage_attestation","attest_author":"https://pith.science/pith/IUJUTK3PXTV5TZBR5RAW432HST/action/author_attestation","sign_citation":"https://pith.science/pith/IUJUTK3PXTV5TZBR5RAW432HST/action/citation_signature","submit_replication":"https://pith.science/pith/IUJUTK3PXTV5TZBR5RAW432HST/action/replication_record"}},"created_at":"2026-07-05T10:18:07.805877+00:00","updated_at":"2026-07-05T10:18:07.805877+00:00"}