{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:JOMRU4EDVT5KNMZVROKAXVSN7M","short_pith_number":"pith:JOMRU4ED","schema_version":"1.0","canonical_sha256":"4b991a7083acfaa6b3358b940bd64dfb3fd2ba581617135d80c67c204ba6243f","source":{"kind":"arxiv","id":"2406.01288","version":2},"attestation_state":"computed","paper":{"title":"Improved Few-Shot Jailbreaking Can Circumvent Aligned Language Models and Their Defenses","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chao Du, Jing Jiang, Min Lin, Qian Liu, Tianyu Pang, Xiaosen Zheng","submitted_at":"2024-06-03T12:59:17Z","abstract_excerpt":"Recently, Anil et al. (2024) show that many-shot (up to hundreds of) demonstrations can jailbreak state-of-the-art LLMs by exploiting their long-context capability. Nevertheless, is it possible to use few-shot demonstrations to efficiently jailbreak LLMs within limited context sizes? While the vanilla few-shot jailbreaking may be inefficient, we propose improved techniques such as injecting special system tokens like [/INST] and employing demo-level random search from a collected demo pool. These simple techniques result in surprisingly effective jailbreaking against aligned LLMs (even with ad"},"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":"2406.01288","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-03T12:59:17Z","cross_cats_sorted":["cs.AI","cs.CR","cs.LG"],"title_canon_sha256":"f5cb188eb0d926d1f89a581abcc7cb0ddc7f1a97649cf8a85bbfd2c5ecd507a8","abstract_canon_sha256":"9e88f7daa79d0c81bd9c087887a4dc12de43b7af4fe9b13ae5802c71190663e9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:28:31.744904Z","signature_b64":"ZT257j28Jy5Tabt401aLVY7c4hcMMNVC04jD2cqzA92VFoK/CmlBZ6O4Gjhfha1+LjmD8ii0IEMTfKJGREo/Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4b991a7083acfaa6b3358b940bd64dfb3fd2ba581617135d80c67c204ba6243f","last_reissued_at":"2026-07-05T09:28:31.744342Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:28:31.744342Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improved Few-Shot Jailbreaking Can Circumvent Aligned Language Models and Their Defenses","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chao Du, Jing Jiang, Min Lin, Qian Liu, Tianyu Pang, Xiaosen Zheng","submitted_at":"2024-06-03T12:59:17Z","abstract_excerpt":"Recently, Anil et al. (2024) show that many-shot (up to hundreds of) demonstrations can jailbreak state-of-the-art LLMs by exploiting their long-context capability. Nevertheless, is it possible to use few-shot demonstrations to efficiently jailbreak LLMs within limited context sizes? While the vanilla few-shot jailbreaking may be inefficient, we propose improved techniques such as injecting special system tokens like [/INST] and employing demo-level random search from a collected demo pool. These simple techniques result in surprisingly effective jailbreaking against aligned LLMs (even with ad"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.01288","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/2406.01288/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":"2406.01288","created_at":"2026-07-05T09:28:31.744417+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.01288v2","created_at":"2026-07-05T09:28:31.744417+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.01288","created_at":"2026-07-05T09:28:31.744417+00:00"},{"alias_kind":"pith_short_12","alias_value":"JOMRU4EDVT5K","created_at":"2026-07-05T09:28:31.744417+00:00"},{"alias_kind":"pith_short_16","alias_value":"JOMRU4EDVT5KNMZV","created_at":"2026-07-05T09:28:31.744417+00:00"},{"alias_kind":"pith_short_8","alias_value":"JOMRU4ED","created_at":"2026-07-05T09:28:31.744417+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25476","citing_title":"A Red Teaming Framework for Large Language Models: A Case Study on Faithfulness Evaluation","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2502.05206","citing_title":"Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety","ref_index":96,"is_internal_anchor":false},{"citing_arxiv_id":"2404.01318","citing_title":"JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2407.04295","citing_title":"Jailbreak Attacks and Defenses Against Large Language Models: A Survey","ref_index":120,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JOMRU4EDVT5KNMZVROKAXVSN7M","json":"https://pith.science/pith/JOMRU4EDVT5KNMZVROKAXVSN7M.json","graph_json":"https://pith.science/api/pith-number/JOMRU4EDVT5KNMZVROKAXVSN7M/graph.json","events_json":"https://pith.science/api/pith-number/JOMRU4EDVT5KNMZVROKAXVSN7M/events.json","paper":"https://pith.science/paper/JOMRU4ED"},"agent_actions":{"view_html":"https://pith.science/pith/JOMRU4EDVT5KNMZVROKAXVSN7M","download_json":"https://pith.science/pith/JOMRU4EDVT5KNMZVROKAXVSN7M.json","view_paper":"https://pith.science/paper/JOMRU4ED","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.01288&json=true","fetch_graph":"https://pith.science/api/pith-number/JOMRU4EDVT5KNMZVROKAXVSN7M/graph.json","fetch_events":"https://pith.science/api/pith-number/JOMRU4EDVT5KNMZVROKAXVSN7M/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JOMRU4EDVT5KNMZVROKAXVSN7M/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JOMRU4EDVT5KNMZVROKAXVSN7M/action/storage_attestation","attest_author":"https://pith.science/pith/JOMRU4EDVT5KNMZVROKAXVSN7M/action/author_attestation","sign_citation":"https://pith.science/pith/JOMRU4EDVT5KNMZVROKAXVSN7M/action/citation_signature","submit_replication":"https://pith.science/pith/JOMRU4EDVT5KNMZVROKAXVSN7M/action/replication_record"}},"created_at":"2026-07-05T09:28:31.744417+00:00","updated_at":"2026-07-05T09:28:31.744417+00:00"}