{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:CVLRJOVBCAIHHDEHP2O6LAHP5K","short_pith_number":"pith:CVLRJOVB","schema_version":"1.0","canonical_sha256":"155714baa11010738c877e9de580efea84282014d58bf1ccdc8d0f92efa5fc3c","source":{"kind":"arxiv","id":"2504.03174","version":1},"attestation_state":"computed","paper":{"title":"Multi-lingual Multi-turn Automated Red Teaming for LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Abhishek Singhania, Amani Namboori, Christophe Dupuy, Shivam Mangale","submitted_at":"2025-04-04T05:06:12Z","abstract_excerpt":"Language Model Models (LLMs) have improved dramatically in the past few years, increasing their adoption and the scope of their capabilities over time. A significant amount of work is dedicated to ``model alignment'', i.e., preventing LLMs to generate unsafe responses when deployed into customer-facing applications. One popular method to evaluate safety risks is \\textit{red-teaming}, where agents attempt to bypass alignment by crafting elaborate prompts that trigger unsafe responses from a model. Standard human-driven red-teaming is costly, time-consuming and rarely covers all the recent featu"},"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":"2504.03174","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-04T05:06:12Z","cross_cats_sorted":[],"title_canon_sha256":"3dbea47a89a96063dc8136accf66718bcb6614c18e623035e9c39b3c35469f7a","abstract_canon_sha256":"b931d559520e891fd8b310374648ed3ef6662d3db6b44d4f6105e44b5ba8bc86"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:44:27.578578Z","signature_b64":"F240o1h7JfWIGFIy3ITOGcwJNTHPYu4fTUX6FlJzGOfDZ3wrZ6ydQ23wOiumFRjaskEKWByy4g1AhsjOIVbwCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"155714baa11010738c877e9de580efea84282014d58bf1ccdc8d0f92efa5fc3c","last_reissued_at":"2026-07-05T10:44:27.578096Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:44:27.578096Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-lingual Multi-turn Automated Red Teaming for LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Abhishek Singhania, Amani Namboori, Christophe Dupuy, Shivam Mangale","submitted_at":"2025-04-04T05:06:12Z","abstract_excerpt":"Language Model Models (LLMs) have improved dramatically in the past few years, increasing their adoption and the scope of their capabilities over time. A significant amount of work is dedicated to ``model alignment'', i.e., preventing LLMs to generate unsafe responses when deployed into customer-facing applications. One popular method to evaluate safety risks is \\textit{red-teaming}, where agents attempt to bypass alignment by crafting elaborate prompts that trigger unsafe responses from a model. Standard human-driven red-teaming is costly, time-consuming and rarely covers all the recent featu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.03174","kind":"arxiv","version":1},"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/2504.03174/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":"2504.03174","created_at":"2026-07-05T10:44:27.578153+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.03174v1","created_at":"2026-07-05T10:44:27.578153+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.03174","created_at":"2026-07-05T10:44:27.578153+00:00"},{"alias_kind":"pith_short_12","alias_value":"CVLRJOVBCAIH","created_at":"2026-07-05T10:44:27.578153+00:00"},{"alias_kind":"pith_short_16","alias_value":"CVLRJOVBCAIHHDEH","created_at":"2026-07-05T10:44:27.578153+00:00"},{"alias_kind":"pith_short_8","alias_value":"CVLRJOVB","created_at":"2026-07-05T10:44:27.578153+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.18988","citing_title":"Surviving the Unseen: Predictive Defense for Novel Multi-Turn Multimodal Attacks","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CVLRJOVBCAIHHDEHP2O6LAHP5K","json":"https://pith.science/pith/CVLRJOVBCAIHHDEHP2O6LAHP5K.json","graph_json":"https://pith.science/api/pith-number/CVLRJOVBCAIHHDEHP2O6LAHP5K/graph.json","events_json":"https://pith.science/api/pith-number/CVLRJOVBCAIHHDEHP2O6LAHP5K/events.json","paper":"https://pith.science/paper/CVLRJOVB"},"agent_actions":{"view_html":"https://pith.science/pith/CVLRJOVBCAIHHDEHP2O6LAHP5K","download_json":"https://pith.science/pith/CVLRJOVBCAIHHDEHP2O6LAHP5K.json","view_paper":"https://pith.science/paper/CVLRJOVB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.03174&json=true","fetch_graph":"https://pith.science/api/pith-number/CVLRJOVBCAIHHDEHP2O6LAHP5K/graph.json","fetch_events":"https://pith.science/api/pith-number/CVLRJOVBCAIHHDEHP2O6LAHP5K/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CVLRJOVBCAIHHDEHP2O6LAHP5K/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CVLRJOVBCAIHHDEHP2O6LAHP5K/action/storage_attestation","attest_author":"https://pith.science/pith/CVLRJOVBCAIHHDEHP2O6LAHP5K/action/author_attestation","sign_citation":"https://pith.science/pith/CVLRJOVBCAIHHDEHP2O6LAHP5K/action/citation_signature","submit_replication":"https://pith.science/pith/CVLRJOVBCAIHHDEHP2O6LAHP5K/action/replication_record"}},"created_at":"2026-07-05T10:44:27.578153+00:00","updated_at":"2026-07-05T10:44:27.578153+00:00"}