{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RYOQBQZMGVDAVFYYZGLZHXFQB5","short_pith_number":"pith:RYOQBQZM","schema_version":"1.0","canonical_sha256":"8e1d00c32c35460a9718c99793dcb00f740757ed74455b3b91a29c05f72bedec","source":{"kind":"arxiv","id":"2405.21018","version":2},"attestation_state":"computed","paper":{"title":"Improved Techniques for Optimization-Based Jailbreaking on Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Chao Du, Jindong Gu, Min Lin, Tianyu Pang, Xiaochun Cao, Xiaojun Jia, Yang Liu, Yihao Huang","submitted_at":"2024-05-31T17:07:15Z","abstract_excerpt":"Large language models (LLMs) are being rapidly developed, and a key component of their widespread deployment is their safety-related alignment. Many red-teaming efforts aim to jailbreak LLMs, where among these efforts, the Greedy Coordinate Gradient (GCG) attack's success has led to a growing interest in the study of optimization-based jailbreaking techniques. Although GCG is a significant milestone, its attacking efficiency remains unsatisfactory. In this paper, we present several improved (empirical) techniques for optimization-based jailbreaks like GCG. We first observe that the single targ"},"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":"2405.21018","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-05-31T17:07:15Z","cross_cats_sorted":["cs.CL","cs.CR"],"title_canon_sha256":"b63df225b51d4ac137d477efd0bb3bec26554399c14c489bb0ce880c465eb9ac","abstract_canon_sha256":"e0f2c1f10376b9a20a611752ac10f86d3932157ad7b4b3eef09209a0ad62d356"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:27:49.442204Z","signature_b64":"sZzkl8Ctd7tEuflR/etPsWaVYjRkY/bNVPhX4PNsuRB5aQqyDWQytkXfkmTqIy6b0zgF6FgrN1PGrCdzP7XZAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8e1d00c32c35460a9718c99793dcb00f740757ed74455b3b91a29c05f72bedec","last_reissued_at":"2026-07-05T08:27:49.441697Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:27:49.441697Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improved Techniques for Optimization-Based Jailbreaking on Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Chao Du, Jindong Gu, Min Lin, Tianyu Pang, Xiaochun Cao, Xiaojun Jia, Yang Liu, Yihao Huang","submitted_at":"2024-05-31T17:07:15Z","abstract_excerpt":"Large language models (LLMs) are being rapidly developed, and a key component of their widespread deployment is their safety-related alignment. Many red-teaming efforts aim to jailbreak LLMs, where among these efforts, the Greedy Coordinate Gradient (GCG) attack's success has led to a growing interest in the study of optimization-based jailbreaking techniques. Although GCG is a significant milestone, its attacking efficiency remains unsatisfactory. In this paper, we present several improved (empirical) techniques for optimization-based jailbreaks like GCG. We first observe that the single targ"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.21018","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/2405.21018/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":"2405.21018","created_at":"2026-07-05T08:27:49.441762+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.21018v2","created_at":"2026-07-05T08:27:49.441762+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.21018","created_at":"2026-07-05T08:27:49.441762+00:00"},{"alias_kind":"pith_short_12","alias_value":"RYOQBQZMGVDA","created_at":"2026-07-05T08:27:49.441762+00:00"},{"alias_kind":"pith_short_16","alias_value":"RYOQBQZMGVDAVFYY","created_at":"2026-07-05T08:27:49.441762+00:00"},{"alias_kind":"pith_short_8","alias_value":"RYOQBQZM","created_at":"2026-07-05T08:27:49.441762+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":19,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2607.08180","citing_title":"Out of Sight: Compression-Aware Content Protection against Agentic Crawlers","ref_index":46,"is_internal_anchor":true},{"citing_arxiv_id":"2607.07461","citing_title":"Mitigating Taint-Style Vulnerabilities in MCP Servers via Security-Aware Tool Descriptions","ref_index":17,"is_internal_anchor":true},{"citing_arxiv_id":"2606.25476","citing_title":"A Red Teaming Framework for Large Language Models: A Case Study on Faithfulness Evaluation","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2606.19887","citing_title":"FinRED: An Expert-Guided Benchmark Generation and Evaluation Framework for Financial LLM Red-Teaming","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2606.15531","citing_title":"Greedy Coordinate Diffusion: Effective and Semantically Coherent Adversarial Attacks via Diffusion Guidance","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05609","citing_title":"SlotGCG: Exploiting the Positional Vulnerability in LLMs for Jailbreak Attacks","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29602","citing_title":"CogniVerse: Revolutionizing Multi-Modal Retrieval-Augmented Generation with Cognitive Reflection and Geometric Reasoning","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2410.15362","citing_title":"Faster-GCG: Efficient Discrete Optimization Jailbreak Attacks against Aligned Large Language Models","ref_index":4,"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":94,"is_internal_anchor":false},{"citing_arxiv_id":"2503.02574","citing_title":"LLM-Safety Evaluations Lack Robustness","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2505.11548","citing_title":"One Shot Dominance: Knowledge Poisoning Attack on Retrieval-Augmented Generation Systems","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21674","citing_title":"Adversarial Reframing: A Framework for Targeted Generation in Language Models","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2602.14211","citing_title":"SkillJect: Effectively Automating Skill-Based Prompt Injection for Skill-Enabled Agents","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19790","citing_title":"Hidden Reliability Risks in Large Language Models: Systematic Identification of Precision-Induced Output Disagreements","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24082","citing_title":"Jailbreaking Frontier Foundation Models Through Intention Deception","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00236","citing_title":"Attention Is Where You Attack","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10326","citing_title":"Jailbreaking the Matrix: Nullspace Steering for Controlled Model Subversion","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05498","citing_title":"JailWAM: Jailbreaking World Action Models in Robot Control","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15717","citing_title":"Into the Gray Zone: Domain Contexts Can Blur LLM Safety Boundaries","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RYOQBQZMGVDAVFYYZGLZHXFQB5","json":"https://pith.science/pith/RYOQBQZMGVDAVFYYZGLZHXFQB5.json","graph_json":"https://pith.science/api/pith-number/RYOQBQZMGVDAVFYYZGLZHXFQB5/graph.json","events_json":"https://pith.science/api/pith-number/RYOQBQZMGVDAVFYYZGLZHXFQB5/events.json","paper":"https://pith.science/paper/RYOQBQZM"},"agent_actions":{"view_html":"https://pith.science/pith/RYOQBQZMGVDAVFYYZGLZHXFQB5","download_json":"https://pith.science/pith/RYOQBQZMGVDAVFYYZGLZHXFQB5.json","view_paper":"https://pith.science/paper/RYOQBQZM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.21018&json=true","fetch_graph":"https://pith.science/api/pith-number/RYOQBQZMGVDAVFYYZGLZHXFQB5/graph.json","fetch_events":"https://pith.science/api/pith-number/RYOQBQZMGVDAVFYYZGLZHXFQB5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RYOQBQZMGVDAVFYYZGLZHXFQB5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RYOQBQZMGVDAVFYYZGLZHXFQB5/action/storage_attestation","attest_author":"https://pith.science/pith/RYOQBQZMGVDAVFYYZGLZHXFQB5/action/author_attestation","sign_citation":"https://pith.science/pith/RYOQBQZMGVDAVFYYZGLZHXFQB5/action/citation_signature","submit_replication":"https://pith.science/pith/RYOQBQZMGVDAVFYYZGLZHXFQB5/action/replication_record"}},"created_at":"2026-07-05T08:27:49.441762+00:00","updated_at":"2026-07-05T08:27:49.441762+00:00"}