{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:22SJEIHKQ2ZDBKV4SMXIMLFWNI","short_pith_number":"pith:22SJEIHK","schema_version":"1.0","canonical_sha256":"d6a49220ea86b230aabc932e862cb66a2835835cb4b5df682e94e01fb7956930","source":{"kind":"arxiv","id":"2503.20084","version":2},"attestation_state":"computed","paper":{"title":"Can Multi-modal (reasoning) LLMs work as deepfake detectors?","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Hengwei Xu, Kidus Zewde, Ning-Yau Cheng, Qinzhe Liu, Simiao Ren, Tsang (Dennis) Ng, Xiaoou Zhan, Yao Yao, Yifei Chen, Zisheng Liang","submitted_at":"2025-03-25T21:47:29Z","abstract_excerpt":"Deepfake detection remains a critical challenge in the era of advanced generative models, particularly as synthetic media becomes more sophisticated. In this study, we explore the potential of state of the art multi-modal (reasoning) large language models (LLMs) for deepfake image detection such as (OpenAI O1/4o, Gemini thinking Flash 2, Deepseek Janus, Grok 3, llama 3.2, Qwen 2/2.5 VL, Mistral Pixtral, Claude 3.5/3.7 sonnet) . We benchmark 12 latest multi-modal LLMs against traditional deepfake detection methods across multiple datasets, including recently published real-world deepfake imager"},"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":"2503.20084","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-25T21:47:29Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"374fbdffec138b5e5bae4c649f6aaeb7a60d469f9cc2547e08109b779f4c2d31","abstract_canon_sha256":"59f9b9e7c535db310ae6a24b733f7acd600a29509e080ede25501d072246482a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:41:23.269949Z","signature_b64":"QGiAfHIqqYsmKJgy/SEOooiHnMe79b5emkbnGrbSCnvvow7NHdRNvgVwptWPRohSiP3I6MQXuRh1BIzmAE1OCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d6a49220ea86b230aabc932e862cb66a2835835cb4b5df682e94e01fb7956930","last_reissued_at":"2026-07-05T10:41:23.269422Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:41:23.269422Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Can Multi-modal (reasoning) LLMs work as deepfake detectors?","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Hengwei Xu, Kidus Zewde, Ning-Yau Cheng, Qinzhe Liu, Simiao Ren, Tsang (Dennis) Ng, Xiaoou Zhan, Yao Yao, Yifei Chen, Zisheng Liang","submitted_at":"2025-03-25T21:47:29Z","abstract_excerpt":"Deepfake detection remains a critical challenge in the era of advanced generative models, particularly as synthetic media becomes more sophisticated. In this study, we explore the potential of state of the art multi-modal (reasoning) large language models (LLMs) for deepfake image detection such as (OpenAI O1/4o, Gemini thinking Flash 2, Deepseek Janus, Grok 3, llama 3.2, Qwen 2/2.5 VL, Mistral Pixtral, Claude 3.5/3.7 sonnet) . We benchmark 12 latest multi-modal LLMs against traditional deepfake detection methods across multiple datasets, including recently published real-world deepfake imager"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.20084","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/2503.20084/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":"2503.20084","created_at":"2026-07-05T10:41:23.269487+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.20084v2","created_at":"2026-07-05T10:41:23.269487+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.20084","created_at":"2026-07-05T10:41:23.269487+00:00"},{"alias_kind":"pith_short_12","alias_value":"22SJEIHKQ2ZD","created_at":"2026-07-05T10:41:23.269487+00:00"},{"alias_kind":"pith_short_16","alias_value":"22SJEIHKQ2ZDBKV4","created_at":"2026-07-05T10:41:23.269487+00:00"},{"alias_kind":"pith_short_8","alias_value":"22SJEIHK","created_at":"2026-07-05T10:41:23.269487+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.25212","citing_title":"Noncrossing Duality and the Geometry of Positive Tropical Linear Spaces","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25213","citing_title":"When the Forger Is the Judge: GPT-Image-2 Cannot Recognize Its Own Faked Documents","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25370","citing_title":"GPT-Image-2 in the Wild: A Twitter Dataset of Self-Reported AI-Generated Images from the First Week of Deployment","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25370","citing_title":"GPT-Image-2 in the Wild: A Twitter Dataset of Self-Reported AI-Generated Images from the First Week of Deployment","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05475","citing_title":"A Synthetic Eye Movement Dataset for Script Reading Detection: Real Trajectory Replay on a 3D Simulator","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/22SJEIHKQ2ZDBKV4SMXIMLFWNI","json":"https://pith.science/pith/22SJEIHKQ2ZDBKV4SMXIMLFWNI.json","graph_json":"https://pith.science/api/pith-number/22SJEIHKQ2ZDBKV4SMXIMLFWNI/graph.json","events_json":"https://pith.science/api/pith-number/22SJEIHKQ2ZDBKV4SMXIMLFWNI/events.json","paper":"https://pith.science/paper/22SJEIHK"},"agent_actions":{"view_html":"https://pith.science/pith/22SJEIHKQ2ZDBKV4SMXIMLFWNI","download_json":"https://pith.science/pith/22SJEIHKQ2ZDBKV4SMXIMLFWNI.json","view_paper":"https://pith.science/paper/22SJEIHK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.20084&json=true","fetch_graph":"https://pith.science/api/pith-number/22SJEIHKQ2ZDBKV4SMXIMLFWNI/graph.json","fetch_events":"https://pith.science/api/pith-number/22SJEIHKQ2ZDBKV4SMXIMLFWNI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/22SJEIHKQ2ZDBKV4SMXIMLFWNI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/22SJEIHKQ2ZDBKV4SMXIMLFWNI/action/storage_attestation","attest_author":"https://pith.science/pith/22SJEIHKQ2ZDBKV4SMXIMLFWNI/action/author_attestation","sign_citation":"https://pith.science/pith/22SJEIHKQ2ZDBKV4SMXIMLFWNI/action/citation_signature","submit_replication":"https://pith.science/pith/22SJEIHKQ2ZDBKV4SMXIMLFWNI/action/replication_record"}},"created_at":"2026-07-05T10:41:23.269487+00:00","updated_at":"2026-07-05T10:41:23.269487+00:00"}