{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:F6D7RO3DNUWQTYV6EG6APWYBRB","short_pith_number":"pith:F6D7RO3D","schema_version":"1.0","canonical_sha256":"2f87f8bb636d2d09e2be21bc07db01887ae4bbe938817bf721f2fa7b7763bf5f","source":{"kind":"arxiv","id":"2503.12874","version":2},"attestation_state":"computed","paper":{"title":"Evolution-based Region Adversarial Prompt Learning for Robustness Enhancement in Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ke Ma, Sensen Gao, Simeng Qin, Wei Dong, Xiaochun Cao, Xiaojun Jia, Xinfeng Li, Yang Liu, Yihao Huang","submitted_at":"2025-03-17T07:08:47Z","abstract_excerpt":"Large pre-trained vision-language models (VLMs), such as CLIP, demonstrate impressive generalization but remain highly vulnerable to adversarial examples (AEs). Previous work has explored robust text prompts through adversarial training, achieving some improvement in both robustness and generalization. However, they primarily rely on singlegradient direction perturbations (e.g., PGD) to generate AEs, which lack diversity, resulting in limited improvement in adversarial robustness. To address these limitations, we propose an evolution-based region adversarial prompt tuning method called ER-APT,"},"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":"2503.12874","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-03-17T07:08:47Z","cross_cats_sorted":[],"title_canon_sha256":"d0d467ab30a6e9e71ed6bc5864f46599974014c675ddff4fd71f1e7f78e6b632","abstract_canon_sha256":"ccd7d40d424cff24328dc8881dbc9cd778d1929af7164a291b54fc8c0d227cc3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:33:10.096351Z","signature_b64":"YfCLxsyZGE4GUff0fRRrrfvCE70Y5LwRJaqU/hoBGEmX2pY4qwS4nwVIT8Cz8gHO9S78hxltgSAqnMBMOLzIDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2f87f8bb636d2d09e2be21bc07db01887ae4bbe938817bf721f2fa7b7763bf5f","last_reissued_at":"2026-07-05T10:33:10.095756Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:33:10.095756Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Evolution-based Region Adversarial Prompt Learning for Robustness Enhancement in Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ke Ma, Sensen Gao, Simeng Qin, Wei Dong, Xiaochun Cao, Xiaojun Jia, Xinfeng Li, Yang Liu, Yihao Huang","submitted_at":"2025-03-17T07:08:47Z","abstract_excerpt":"Large pre-trained vision-language models (VLMs), such as CLIP, demonstrate impressive generalization but remain highly vulnerable to adversarial examples (AEs). Previous work has explored robust text prompts through adversarial training, achieving some improvement in both robustness and generalization. However, they primarily rely on singlegradient direction perturbations (e.g., PGD) to generate AEs, which lack diversity, resulting in limited improvement in adversarial robustness. To address these limitations, we propose an evolution-based region adversarial prompt tuning method called ER-APT,"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.12874","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/2503.12874/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":"2503.12874","created_at":"2026-07-05T10:33:10.095825+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.12874v2","created_at":"2026-07-05T10:33:10.095825+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.12874","created_at":"2026-07-05T10:33:10.095825+00:00"},{"alias_kind":"pith_short_12","alias_value":"F6D7RO3DNUWQ","created_at":"2026-07-05T10:33:10.095825+00:00"},{"alias_kind":"pith_short_16","alias_value":"F6D7RO3DNUWQTYV6","created_at":"2026-07-05T10:33:10.095825+00:00"},{"alias_kind":"pith_short_8","alias_value":"F6D7RO3D","created_at":"2026-07-05T10:33:10.095825+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.27823","citing_title":"Disentangling Adversarial Prompts: A Semantic-Graph Defense for Robust LLM Security","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29602","citing_title":"CogniVerse: Revolutionizing Multi-Modal Retrieval-Augmented Generation with Cognitive Reflection and Geometric Reasoning","ref_index":61,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/F6D7RO3DNUWQTYV6EG6APWYBRB","json":"https://pith.science/pith/F6D7RO3DNUWQTYV6EG6APWYBRB.json","graph_json":"https://pith.science/api/pith-number/F6D7RO3DNUWQTYV6EG6APWYBRB/graph.json","events_json":"https://pith.science/api/pith-number/F6D7RO3DNUWQTYV6EG6APWYBRB/events.json","paper":"https://pith.science/paper/F6D7RO3D"},"agent_actions":{"view_html":"https://pith.science/pith/F6D7RO3DNUWQTYV6EG6APWYBRB","download_json":"https://pith.science/pith/F6D7RO3DNUWQTYV6EG6APWYBRB.json","view_paper":"https://pith.science/paper/F6D7RO3D","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.12874&json=true","fetch_graph":"https://pith.science/api/pith-number/F6D7RO3DNUWQTYV6EG6APWYBRB/graph.json","fetch_events":"https://pith.science/api/pith-number/F6D7RO3DNUWQTYV6EG6APWYBRB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F6D7RO3DNUWQTYV6EG6APWYBRB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F6D7RO3DNUWQTYV6EG6APWYBRB/action/storage_attestation","attest_author":"https://pith.science/pith/F6D7RO3DNUWQTYV6EG6APWYBRB/action/author_attestation","sign_citation":"https://pith.science/pith/F6D7RO3DNUWQTYV6EG6APWYBRB/action/citation_signature","submit_replication":"https://pith.science/pith/F6D7RO3DNUWQTYV6EG6APWYBRB/action/replication_record"}},"created_at":"2026-07-05T10:33:10.095825+00:00","updated_at":"2026-07-05T10:33:10.095825+00:00"}