{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WXK5L3YW3WXEIWLBEDMGRV3AI2","short_pith_number":"pith:WXK5L3YW","schema_version":"1.0","canonical_sha256":"b5d5d5ef16ddae44596120d868d76046afea494c803f1d23b42844ef9bb5db5e","source":{"kind":"arxiv","id":"2503.19683","version":2},"attestation_state":"computed","paper":{"title":"Unlocking the Hidden Potential of CLIP in Generalizable Deepfake Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Andrii Yermakov, Jan Cech, Jiri Matas","submitted_at":"2025-03-25T14:10:54Z","abstract_excerpt":"This paper tackles the challenge of detecting partially manipulated facial deepfakes, which involve subtle alterations to specific facial features while retaining the overall context, posing a greater detection difficulty than fully synthetic faces. We leverage the Contrastive Language-Image Pre-training (CLIP) model, specifically its ViT-L/14 visual encoder, to develop a generalizable detection method that performs robustly across diverse datasets and unknown forgery techniques with minimal modifications to the original model. The proposed approach utilizes parameter-efficient fine-tuning (PE"},"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":"2503.19683","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-25T14:10:54Z","cross_cats_sorted":[],"title_canon_sha256":"affcfcecfefd9c8b65387d44b944b193f98f91ea7c3db7e870d25bf272ec68ea","abstract_canon_sha256":"b338a2170072f4538ed0d58838d3d7ca2374b9cca6d64b2cb8b024e93cbf8b7f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:39:28.617041Z","signature_b64":"fNrG+z/AOgpuo1C3dkc2aNwXwYFukzr0aVhy94OMkM2/lIuduYz6Qfs0HJ265oNmtnKb5ZDaYSA+spi3E0fvCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b5d5d5ef16ddae44596120d868d76046afea494c803f1d23b42844ef9bb5db5e","last_reissued_at":"2026-07-05T10:39:28.616566Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:39:28.616566Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unlocking the Hidden Potential of CLIP in Generalizable Deepfake Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Andrii Yermakov, Jan Cech, Jiri Matas","submitted_at":"2025-03-25T14:10:54Z","abstract_excerpt":"This paper tackles the challenge of detecting partially manipulated facial deepfakes, which involve subtle alterations to specific facial features while retaining the overall context, posing a greater detection difficulty than fully synthetic faces. We leverage the Contrastive Language-Image Pre-training (CLIP) model, specifically its ViT-L/14 visual encoder, to develop a generalizable detection method that performs robustly across diverse datasets and unknown forgery techniques with minimal modifications to the original model. The proposed approach utilizes parameter-efficient fine-tuning (PE"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.19683","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/2503.19683/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":"2503.19683","created_at":"2026-07-05T10:39:28.616624+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.19683v2","created_at":"2026-07-05T10:39:28.616624+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.19683","created_at":"2026-07-05T10:39:28.616624+00:00"},{"alias_kind":"pith_short_12","alias_value":"WXK5L3YW3WXE","created_at":"2026-07-05T10:39:28.616624+00:00"},{"alias_kind":"pith_short_16","alias_value":"WXK5L3YW3WXEIWLB","created_at":"2026-07-05T10:39:28.616624+00:00"},{"alias_kind":"pith_short_8","alias_value":"WXK5L3YW","created_at":"2026-07-05T10:39:28.616624+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07216","citing_title":"Why Fake ? Unveiling the Semantic Vocabulary of Deepfake Detectors","ref_index":43,"is_internal_anchor":true},{"citing_arxiv_id":"2606.01843","citing_title":"Suppressing Forgery-Specific Shortcuts for Generalizable Deepfake Detection","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30528","citing_title":"$\\mu$Flow: Leveraging Average Images for Improving Generalisation of Deepfake Faces Detectors","ref_index":64,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WXK5L3YW3WXEIWLBEDMGRV3AI2","json":"https://pith.science/pith/WXK5L3YW3WXEIWLBEDMGRV3AI2.json","graph_json":"https://pith.science/api/pith-number/WXK5L3YW3WXEIWLBEDMGRV3AI2/graph.json","events_json":"https://pith.science/api/pith-number/WXK5L3YW3WXEIWLBEDMGRV3AI2/events.json","paper":"https://pith.science/paper/WXK5L3YW"},"agent_actions":{"view_html":"https://pith.science/pith/WXK5L3YW3WXEIWLBEDMGRV3AI2","download_json":"https://pith.science/pith/WXK5L3YW3WXEIWLBEDMGRV3AI2.json","view_paper":"https://pith.science/paper/WXK5L3YW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.19683&json=true","fetch_graph":"https://pith.science/api/pith-number/WXK5L3YW3WXEIWLBEDMGRV3AI2/graph.json","fetch_events":"https://pith.science/api/pith-number/WXK5L3YW3WXEIWLBEDMGRV3AI2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WXK5L3YW3WXEIWLBEDMGRV3AI2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WXK5L3YW3WXEIWLBEDMGRV3AI2/action/storage_attestation","attest_author":"https://pith.science/pith/WXK5L3YW3WXEIWLBEDMGRV3AI2/action/author_attestation","sign_citation":"https://pith.science/pith/WXK5L3YW3WXEIWLBEDMGRV3AI2/action/citation_signature","submit_replication":"https://pith.science/pith/WXK5L3YW3WXEIWLBEDMGRV3AI2/action/replication_record"}},"created_at":"2026-07-05T10:39:28.616624+00:00","updated_at":"2026-07-05T10:39:28.616624+00:00"}