{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TC73LDBVDN6LK6BB2DLF5WRPAZ","short_pith_number":"pith:TC73LDBV","schema_version":"1.0","canonical_sha256":"98bfb58c351b7cb57821d0d65eda2f067e9c280ec28f1f72bf20e80cdee0c6d8","source":{"kind":"arxiv","id":"2507.23202","version":1},"attestation_state":"computed","paper":{"title":"Adversarial-Guided Diffusion for Multimodal LLM Attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chengwei Xia, Fan Ma, Kun Zhan, Ruijie Quan, Yi Yang","submitted_at":"2025-07-31T02:57:20Z","abstract_excerpt":"This paper addresses the challenge of generating adversarial image using a diffusion model to deceive multimodal large language models (MLLMs) into generating the targeted responses, while avoiding significant distortion of the clean image. To address the above challenges, we propose an adversarial-guided diffusion (AGD) approach for adversarial attack MLLMs. We introduce adversarial-guided noise to ensure attack efficacy. A key observation in our design is that, unlike most traditional adversarial attacks which embed high-frequency perturbations directly into the clean image, AGD injects 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":"2507.23202","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-07-31T02:57:20Z","cross_cats_sorted":[],"title_canon_sha256":"39c7d82007c730324e5a69fb928190a53dd6130c2a4b9d753d26967637101a57","abstract_canon_sha256":"29baa9fce0e97e76291ba733ac511efa53cd3f73b7e11a4e8c91687951ab3466"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:46:07.359837Z","signature_b64":"B6voy2Iqzq4Mt3nVzMghhNJsRgX4pZOvDA5H/mFL0vUTRhjHjTxD5pgJdnEnE1f5l8jyJ+xkCQ0tv9J40sZEBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"98bfb58c351b7cb57821d0d65eda2f067e9c280ec28f1f72bf20e80cdee0c6d8","last_reissued_at":"2026-07-05T11:46:07.359074Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:46:07.359074Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Adversarial-Guided Diffusion for Multimodal LLM Attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chengwei Xia, Fan Ma, Kun Zhan, Ruijie Quan, Yi Yang","submitted_at":"2025-07-31T02:57:20Z","abstract_excerpt":"This paper addresses the challenge of generating adversarial image using a diffusion model to deceive multimodal large language models (MLLMs) into generating the targeted responses, while avoiding significant distortion of the clean image. To address the above challenges, we propose an adversarial-guided diffusion (AGD) approach for adversarial attack MLLMs. We introduce adversarial-guided noise to ensure attack efficacy. A key observation in our design is that, unlike most traditional adversarial attacks which embed high-frequency perturbations directly into the clean image, AGD injects targ"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.23202","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/2507.23202/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":"2507.23202","created_at":"2026-07-05T11:46:07.359167+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.23202v1","created_at":"2026-07-05T11:46:07.359167+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.23202","created_at":"2026-07-05T11:46:07.359167+00:00"},{"alias_kind":"pith_short_12","alias_value":"TC73LDBVDN6L","created_at":"2026-07-05T11:46:07.359167+00:00"},{"alias_kind":"pith_short_16","alias_value":"TC73LDBVDN6LK6BB","created_at":"2026-07-05T11:46:07.359167+00:00"},{"alias_kind":"pith_short_8","alias_value":"TC73LDBV","created_at":"2026-07-05T11:46:07.359167+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2511.21893","citing_title":"Breaking the Illusion: Consensus-Based Generative Mitigation of Adversarial Illusions in Multi-Modal Embeddings","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TC73LDBVDN6LK6BB2DLF5WRPAZ","json":"https://pith.science/pith/TC73LDBVDN6LK6BB2DLF5WRPAZ.json","graph_json":"https://pith.science/api/pith-number/TC73LDBVDN6LK6BB2DLF5WRPAZ/graph.json","events_json":"https://pith.science/api/pith-number/TC73LDBVDN6LK6BB2DLF5WRPAZ/events.json","paper":"https://pith.science/paper/TC73LDBV"},"agent_actions":{"view_html":"https://pith.science/pith/TC73LDBVDN6LK6BB2DLF5WRPAZ","download_json":"https://pith.science/pith/TC73LDBVDN6LK6BB2DLF5WRPAZ.json","view_paper":"https://pith.science/paper/TC73LDBV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.23202&json=true","fetch_graph":"https://pith.science/api/pith-number/TC73LDBVDN6LK6BB2DLF5WRPAZ/graph.json","fetch_events":"https://pith.science/api/pith-number/TC73LDBVDN6LK6BB2DLF5WRPAZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TC73LDBVDN6LK6BB2DLF5WRPAZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TC73LDBVDN6LK6BB2DLF5WRPAZ/action/storage_attestation","attest_author":"https://pith.science/pith/TC73LDBVDN6LK6BB2DLF5WRPAZ/action/author_attestation","sign_citation":"https://pith.science/pith/TC73LDBVDN6LK6BB2DLF5WRPAZ/action/citation_signature","submit_replication":"https://pith.science/pith/TC73LDBVDN6LK6BB2DLF5WRPAZ/action/replication_record"}},"created_at":"2026-07-05T11:46:07.359167+00:00","updated_at":"2026-07-05T11:46:07.359167+00:00"}