{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:Z6WKJI5SFETOQWYIZWIGLCABA2","short_pith_number":"pith:Z6WKJI5S","schema_version":"1.0","canonical_sha256":"cfaca4a3b22926e85b08cd9065880106a8a2392230c5442fd26da858b3d1e93c","source":{"kind":"arxiv","id":"2410.18927","version":1},"attestation_state":"computed","paper":{"title":"SafeBench: A Safety Evaluation Framework for Multimodal Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CR","authors_text":"Aishan Liu, Dacheng Tao, Jinyang Guo, Lei Huang, Siyuan Liang, Wenbo Zhou, Xianglong Liu, Zonghao Ying","submitted_at":"2024-10-24T17:14:40Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) are showing strong safety concerns (e.g., generating harmful outputs for users), which motivates the development of safety evaluation benchmarks. However, we observe that existing safety benchmarks for MLLMs show limitations in query quality and evaluation reliability limiting the detection of model safety implications as MLLMs continue to evolve. In this paper, we propose \\toolns, a comprehensive framework designed for conducting safety evaluations of MLLMs. Our framework consists of a comprehensive harmful query dataset and an automated evaluation pro"},"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":"2410.18927","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2024-10-24T17:14:40Z","cross_cats_sorted":[],"title_canon_sha256":"5ba11b74c9b4ea23f557373cd2c61a48ae8601cad16e38034120f3d944adbacc","abstract_canon_sha256":"8d90efe6eb03b003b758c59f76146498f477c88e0222d52d9548a065bdee6bf2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:25:22.122858Z","signature_b64":"z476uWAqHsKzE583RdevfDiiFZ9S/fpgrsIILcOmrLU8t6dC7P/uBxDVV90fovrAjWtVbfaCQL+oIqzJaQ7yAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cfaca4a3b22926e85b08cd9065880106a8a2392230c5442fd26da858b3d1e93c","last_reissued_at":"2026-07-05T09:25:22.122350Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:25:22.122350Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SafeBench: A Safety Evaluation Framework for Multimodal Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CR","authors_text":"Aishan Liu, Dacheng Tao, Jinyang Guo, Lei Huang, Siyuan Liang, Wenbo Zhou, Xianglong Liu, Zonghao Ying","submitted_at":"2024-10-24T17:14:40Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) are showing strong safety concerns (e.g., generating harmful outputs for users), which motivates the development of safety evaluation benchmarks. However, we observe that existing safety benchmarks for MLLMs show limitations in query quality and evaluation reliability limiting the detection of model safety implications as MLLMs continue to evolve. In this paper, we propose \\toolns, a comprehensive framework designed for conducting safety evaluations of MLLMs. Our framework consists of a comprehensive harmful query dataset and an automated evaluation pro"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.18927","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/2410.18927/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":"2410.18927","created_at":"2026-07-05T09:25:22.122417+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.18927v1","created_at":"2026-07-05T09:25:22.122417+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.18927","created_at":"2026-07-05T09:25:22.122417+00:00"},{"alias_kind":"pith_short_12","alias_value":"Z6WKJI5SFETO","created_at":"2026-07-05T09:25:22.122417+00:00"},{"alias_kind":"pith_short_16","alias_value":"Z6WKJI5SFETOQWYI","created_at":"2026-07-05T09:25:22.122417+00:00"},{"alias_kind":"pith_short_8","alias_value":"Z6WKJI5S","created_at":"2026-07-05T09:25:22.122417+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18632","citing_title":"ROBOSHACKLES: A Safety Dataset for Human-Injury Prevention in Embodied Foundation Models","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2411.18275","citing_title":"Visual Adversarial Attack on Vision-Language Models for Autonomous Driving","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2507.21540","citing_title":"PRISM: Programmatic Reasoning with Image Sequence Manipulation for LVLM Jailbreaking","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2510.10073","citing_title":"SecureWebArena: A Holistic Security Evaluation Benchmark for LVLM-based Web Agents","ref_index":55,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Z6WKJI5SFETOQWYIZWIGLCABA2","json":"https://pith.science/pith/Z6WKJI5SFETOQWYIZWIGLCABA2.json","graph_json":"https://pith.science/api/pith-number/Z6WKJI5SFETOQWYIZWIGLCABA2/graph.json","events_json":"https://pith.science/api/pith-number/Z6WKJI5SFETOQWYIZWIGLCABA2/events.json","paper":"https://pith.science/paper/Z6WKJI5S"},"agent_actions":{"view_html":"https://pith.science/pith/Z6WKJI5SFETOQWYIZWIGLCABA2","download_json":"https://pith.science/pith/Z6WKJI5SFETOQWYIZWIGLCABA2.json","view_paper":"https://pith.science/paper/Z6WKJI5S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.18927&json=true","fetch_graph":"https://pith.science/api/pith-number/Z6WKJI5SFETOQWYIZWIGLCABA2/graph.json","fetch_events":"https://pith.science/api/pith-number/Z6WKJI5SFETOQWYIZWIGLCABA2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Z6WKJI5SFETOQWYIZWIGLCABA2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Z6WKJI5SFETOQWYIZWIGLCABA2/action/storage_attestation","attest_author":"https://pith.science/pith/Z6WKJI5SFETOQWYIZWIGLCABA2/action/author_attestation","sign_citation":"https://pith.science/pith/Z6WKJI5SFETOQWYIZWIGLCABA2/action/citation_signature","submit_replication":"https://pith.science/pith/Z6WKJI5SFETOQWYIZWIGLCABA2/action/replication_record"}},"created_at":"2026-07-05T09:25:22.122417+00:00","updated_at":"2026-07-05T09:25:22.122417+00:00"}