{"paper":{"title":"AEGIS: A Holistic Benchmark for Evaluating Forensic Analysis of AI-Generated Academic Images","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"AEGIS benchmark reveals that even advanced models detect AI-generated academic images at only 48.80 percent overall accuracy.","cross_cats":["cs.CY"],"primary_cat":"cs.CV","authors_text":"Bo Zhang, Haihong E, Haiyang Sun, Haocheng Gao, Jiacheng Liu, Junpeng Ding, Liangjia Wang, Peilin Gao, Ronghui Xi, Tzu-Yen Ma, Yiling Huang, Yizhuo Zhao, Yuan Liu, Yuanze Li, Yujie Wang, Yuyue Zhang, Zhongjun Yang, Zichen Tang, Zijie Xi, Zirui Wang, Zixin Ding","submitted_at":"2026-04-30T17:56:58Z","abstract_excerpt":"We introduce AEGIS, A holistic benchmark for Evaluating forensic analysis of AI-Generated academic ImageS. Compared to existing benchmarks, AEGIS features three key advances: (1) Domain-Specific Complexity: covering seven academic categories with 39 fine-grained subtypes, exposing intrinsic forensic difficulty, where even GPT-5.1 reaches 48.80% overall performance and expert models achieve only limited localization accuracy (IoU 30.09%); (2) Diverse Forgery Simulations: modeling four prevalent academic forgery strategies across 25 generative models, with 11 yielding average forensic accuracy b"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"AEGIS serves as a diagnostic testbed exposing fundamental limitations in academic image forensics, with even GPT-5.1 reaching only 48.80% overall performance and expert models limited to 30.09% IoU localization accuracy.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The 39 fine-grained academic subtypes and four simulated forgery strategies using 25 generative models sufficiently represent the real-world distribution and difficulty of AI-generated academic images.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"AEGIS benchmark reveals that leading AI models achieve only 48.80% overall accuracy and low localization precision when analyzing AI-generated academic images, exposing gaps between generative and forensic capabilities.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"AEGIS benchmark reveals that even advanced models detect AI-generated academic images at only 48.80 percent overall accuracy.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c4eeda82f238ff179a642da188b5c32885c02ab08bbe6c527d2516d0b1454d57"},"source":{"id":"2604.28177","kind":"arxiv","version":2},"verdict":{"id":"163fd9c5-13d6-4b16-b3e1-cc0d068eb3b1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T07:30:38.284435Z","strongest_claim":"AEGIS serves as a diagnostic testbed exposing fundamental limitations in academic image forensics, with even GPT-5.1 reaching only 48.80% overall performance and expert models limited to 30.09% IoU localization accuracy.","one_line_summary":"AEGIS benchmark reveals that leading AI models achieve only 48.80% overall accuracy and low localization precision when analyzing AI-generated academic images, exposing gaps between generative and forensic capabilities.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The 39 fine-grained academic subtypes and four simulated forgery strategies using 25 generative models sufficiently represent the real-world distribution and difficulty of AI-generated academic images.","pith_extraction_headline":"AEGIS benchmark reveals that even advanced models detect AI-generated academic images at only 48.80 percent overall accuracy."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.28177/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T20:39:31.023454Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T18:33:07.351969Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"a0c1a52141a77c30ce1e620a9dc4c3e2cbaf6e91ffa8789961570ecf4e9ba5ed"},"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"}