{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ZTWL4FFAGLAXSEK2OU2GRDQJJV","short_pith_number":"pith:ZTWL4FFA","schema_version":"1.0","canonical_sha256":"ccecbe14a032c179115a7534688e094d531eaa660b26574ee641dba38fb4748b","source":{"kind":"arxiv","id":"2403.09766","version":1},"attestation_state":"computed","paper":{"title":"An Image Is Worth 1000 Lies: Adversarial Transferability across Prompts on Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fengyuan Liu, Haochen Luo, Jindong Gu, Philip Torr","submitted_at":"2024-03-14T17:59:35Z","abstract_excerpt":"Different from traditional task-specific vision models, recent large VLMs can readily adapt to different vision tasks by simply using different textual instructions, i.e., prompts. However, a well-known concern about traditional task-specific vision models is that they can be misled by imperceptible adversarial perturbations. Furthermore, the concern is exacerbated by the phenomenon that the same adversarial perturbations can fool different task-specific models. Given that VLMs rely on prompts to adapt to different tasks, an intriguing question emerges: Can a single adversarial image mislead a"},"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":"2403.09766","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-03-14T17:59:35Z","cross_cats_sorted":[],"title_canon_sha256":"657ea49bfc90d6bd7d8f39e495c8ad87ca83a88aedf876ed8799ed7bfa6257b0","abstract_canon_sha256":"ac69ede94b79214181e87b2c1a0cb003a0fae3bde61d0183012f7f554af20271"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:56:24.711082Z","signature_b64":"N8ojf5VC2RV+/wV9VNM2bsEXhJMnjWY1HeUrpDUpKUM9rByKGDUEzpiD5XS+uEIe7hOP43/sjD8BWwWIGnUuCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ccecbe14a032c179115a7534688e094d531eaa660b26574ee641dba38fb4748b","last_reissued_at":"2026-07-05T07:56:24.710607Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:56:24.710607Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"An Image Is Worth 1000 Lies: Adversarial Transferability across Prompts on Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fengyuan Liu, Haochen Luo, Jindong Gu, Philip Torr","submitted_at":"2024-03-14T17:59:35Z","abstract_excerpt":"Different from traditional task-specific vision models, recent large VLMs can readily adapt to different vision tasks by simply using different textual instructions, i.e., prompts. However, a well-known concern about traditional task-specific vision models is that they can be misled by imperceptible adversarial perturbations. Furthermore, the concern is exacerbated by the phenomenon that the same adversarial perturbations can fool different task-specific models. Given that VLMs rely on prompts to adapt to different tasks, an intriguing question emerges: Can a single adversarial image mislead a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.09766","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/2403.09766/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":"2403.09766","created_at":"2026-07-05T07:56:24.710667+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.09766v1","created_at":"2026-07-05T07:56:24.710667+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.09766","created_at":"2026-07-05T07:56:24.710667+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZTWL4FFAGLAX","created_at":"2026-07-05T07:56:24.710667+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZTWL4FFAGLAXSEK2","created_at":"2026-07-05T07:56:24.710667+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZTWL4FFA","created_at":"2026-07-05T07:56:24.710667+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07375","citing_title":"On Adversarial Vulnerability of Vision-Language Models through the Lens of Intermediate Spectral Subspaces","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2606.10571","citing_title":"Improving Adversarial Transferability on Vision-Language Pre-training Models via Surrogate-Specific Bias Correction","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28609","citing_title":"JECA^2: Judgment-Explanation Consistent Adversarial Attack against Forensic Vision-Language Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16090","citing_title":"A Cross-Modal Prompt Injection Attack against Large Vision-Language Models with Image-Only Perturbation","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24082","citing_title":"Jailbreaking Frontier Foundation Models Through Intention Deception","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZTWL4FFAGLAXSEK2OU2GRDQJJV","json":"https://pith.science/pith/ZTWL4FFAGLAXSEK2OU2GRDQJJV.json","graph_json":"https://pith.science/api/pith-number/ZTWL4FFAGLAXSEK2OU2GRDQJJV/graph.json","events_json":"https://pith.science/api/pith-number/ZTWL4FFAGLAXSEK2OU2GRDQJJV/events.json","paper":"https://pith.science/paper/ZTWL4FFA"},"agent_actions":{"view_html":"https://pith.science/pith/ZTWL4FFAGLAXSEK2OU2GRDQJJV","download_json":"https://pith.science/pith/ZTWL4FFAGLAXSEK2OU2GRDQJJV.json","view_paper":"https://pith.science/paper/ZTWL4FFA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.09766&json=true","fetch_graph":"https://pith.science/api/pith-number/ZTWL4FFAGLAXSEK2OU2GRDQJJV/graph.json","fetch_events":"https://pith.science/api/pith-number/ZTWL4FFAGLAXSEK2OU2GRDQJJV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZTWL4FFAGLAXSEK2OU2GRDQJJV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZTWL4FFAGLAXSEK2OU2GRDQJJV/action/storage_attestation","attest_author":"https://pith.science/pith/ZTWL4FFAGLAXSEK2OU2GRDQJJV/action/author_attestation","sign_citation":"https://pith.science/pith/ZTWL4FFAGLAXSEK2OU2GRDQJJV/action/citation_signature","submit_replication":"https://pith.science/pith/ZTWL4FFAGLAXSEK2OU2GRDQJJV/action/replication_record"}},"created_at":"2026-07-05T07:56:24.710667+00:00","updated_at":"2026-07-05T07:56:24.710667+00:00"}