{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:O6W7536FQT6C6DK5V6Z3I6UUND","short_pith_number":"pith:O6W7536F","schema_version":"1.0","canonical_sha256":"77adfeefc584fc2f0d5dafb3b47a9468d029489ec408998dfea15c3e838e86ea","source":{"kind":"arxiv","id":"2410.12855","version":2},"attestation_state":"computed","paper":{"title":"JAILJUDGE: A Comprehensive Jailbreak Judge Benchmark with Multi-Agent Enhanced Explanation Evaluation Framework","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dawei Yin, Fan Liu, Hao Liu, Lixin Su, Xinyu Ma, Yue Feng, Zhao Xu","submitted_at":"2024-10-11T14:56:28Z","abstract_excerpt":"Despite advancements in enhancing LLM safety against jailbreak attacks, evaluating LLM defenses remains a challenge, with current methods often lacking explainability and generalization to complex scenarios, leading to incomplete assessments (e.g., direct judgment without reasoning, low F1 score of GPT-4 in complex cases, bias in multilingual scenarios). To address this, we present JAILJUDGE, a comprehensive benchmark featuring diverse risk scenarios, including synthetic, adversarial, in-the-wild, and multilingual prompts, along with high-quality human-annotated datasets. The JAILJUDGE dataset"},"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":"2410.12855","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-11T14:56:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b3a689c44581b1f6b2c82afac2008b3ebd5b8d914a39d7b94ca03c97eaebe082","abstract_canon_sha256":"1ee0a5494f128a15858e7bdcd5f6b6bdf00e90fce01bd43e6600d497c82263fa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:22:23.649403Z","signature_b64":"Huj23jOpNJk4I8MI3YPgKQFBy0F7vtcQuegloKPyX/aUl5oKS32gByKdmzC8EPZCjWGUbnzk/II+KZBK/beZAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"77adfeefc584fc2f0d5dafb3b47a9468d029489ec408998dfea15c3e838e86ea","last_reissued_at":"2026-07-05T09:22:23.648928Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:22:23.648928Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"JAILJUDGE: A Comprehensive Jailbreak Judge Benchmark with Multi-Agent Enhanced Explanation Evaluation Framework","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dawei Yin, Fan Liu, Hao Liu, Lixin Su, Xinyu Ma, Yue Feng, Zhao Xu","submitted_at":"2024-10-11T14:56:28Z","abstract_excerpt":"Despite advancements in enhancing LLM safety against jailbreak attacks, evaluating LLM defenses remains a challenge, with current methods often lacking explainability and generalization to complex scenarios, leading to incomplete assessments (e.g., direct judgment without reasoning, low F1 score of GPT-4 in complex cases, bias in multilingual scenarios). To address this, we present JAILJUDGE, a comprehensive benchmark featuring diverse risk scenarios, including synthetic, adversarial, in-the-wild, and multilingual prompts, along with high-quality human-annotated datasets. The JAILJUDGE dataset"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.12855","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/2410.12855/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":"2410.12855","created_at":"2026-07-05T09:22:23.648983+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.12855v2","created_at":"2026-07-05T09:22:23.648983+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.12855","created_at":"2026-07-05T09:22:23.648983+00:00"},{"alias_kind":"pith_short_12","alias_value":"O6W7536FQT6C","created_at":"2026-07-05T09:22:23.648983+00:00"},{"alias_kind":"pith_short_16","alias_value":"O6W7536FQT6C6DK5","created_at":"2026-07-05T09:22:23.648983+00:00"},{"alias_kind":"pith_short_8","alias_value":"O6W7536F","created_at":"2026-07-05T09:22:23.648983+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"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":32,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07335","citing_title":"Defending Jailbreak Attacks on Large Language Models via Manifold Trajectory Kinetics","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2505.16120","citing_title":"LLM-Powered AI Agent Systems and Their Applications in Industry","ref_index":115,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12746","citing_title":"CoT-Guard: Small Models for Strong Monitoring","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09225","citing_title":"The Art of the Jailbreak: Formulating Jailbreak Attacks for LLM Security Beyond Binary Scoring","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24983","citing_title":"Adaptive Prompt Embedding Optimization for LLM Jailbreaking","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05058","citing_title":"SoK: Robustness in Large Language Models against Jailbreak Attacks","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01034","citing_title":"A Theoretical Game of Attacks via Compositional Skills","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11309","citing_title":"The Salami Slicing Threat: Exploiting Cumulative Risks in LLM Systems","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17301","citing_title":"RoTRAG: Rule of Thumb Reasoning for Conversation Harm Detection with Retrieval-Augmented Generation","ref_index":22,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/O6W7536FQT6C6DK5V6Z3I6UUND","json":"https://pith.science/pith/O6W7536FQT6C6DK5V6Z3I6UUND.json","graph_json":"https://pith.science/api/pith-number/O6W7536FQT6C6DK5V6Z3I6UUND/graph.json","events_json":"https://pith.science/api/pith-number/O6W7536FQT6C6DK5V6Z3I6UUND/events.json","paper":"https://pith.science/paper/O6W7536F"},"agent_actions":{"view_html":"https://pith.science/pith/O6W7536FQT6C6DK5V6Z3I6UUND","download_json":"https://pith.science/pith/O6W7536FQT6C6DK5V6Z3I6UUND.json","view_paper":"https://pith.science/paper/O6W7536F","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.12855&json=true","fetch_graph":"https://pith.science/api/pith-number/O6W7536FQT6C6DK5V6Z3I6UUND/graph.json","fetch_events":"https://pith.science/api/pith-number/O6W7536FQT6C6DK5V6Z3I6UUND/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/O6W7536FQT6C6DK5V6Z3I6UUND/action/timestamp_anchor","attest_storage":"https://pith.science/pith/O6W7536FQT6C6DK5V6Z3I6UUND/action/storage_attestation","attest_author":"https://pith.science/pith/O6W7536FQT6C6DK5V6Z3I6UUND/action/author_attestation","sign_citation":"https://pith.science/pith/O6W7536FQT6C6DK5V6Z3I6UUND/action/citation_signature","submit_replication":"https://pith.science/pith/O6W7536FQT6C6DK5V6Z3I6UUND/action/replication_record"}},"created_at":"2026-07-05T09:22:23.648983+00:00","updated_at":"2026-07-05T09:22:23.648983+00:00"}