{"paper":{"title":"ImageAttributionBench: How Far Are We from Generalizable Attribution?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Existing image attribution methods fail to generalize to unseen semantics and degraded images.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chao Gong, Jingjing Chen, Tingshu Mou, Xingjun Ma, Zhipeng Wei","submitted_at":"2026-05-13T04:01:10Z","abstract_excerpt":"The rapid advancement of generative AI has enabled the creation of highly realistic and diverse synthetic images, posing critical challenges for image provenance and misinformation detection. This underscores the urgent need for effective image attribution. However, existing attribution datasets are constrained by limited scale, outdated generation methods, and insufficient semantic diversity - hindering the development of robust and generalizable attribution models. To address these limitations, we introduce ImageAttributionBench, a comprehensive dataset comprising images synthesized by a wid"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"current methods exhibit consistently poor performance, revealing significant limitations in their robustness and generalization to unseen semantic content.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the two evaluation settings (training on balanced split then testing on degraded images, and training/testing on semantically disjoint splits) adequately simulate real-world attribution scenarios.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ImageAttributionBench is a benchmark dataset demonstrating that state-of-the-art image attribution methods lack robustness to image degradation and fail to generalize to semantically disjoint domains.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Existing image attribution methods fail to generalize to unseen semantics and degraded images.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0d75b5c4428b6502f6089f09c39095c5ef25a2b09feb4cfbc6b5aee21793edfe"},"source":{"id":"2605.12967","kind":"arxiv","version":1},"verdict":{"id":"27d1a145-6887-47b9-be16-1442e7636ce7","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:35:01.327649Z","strongest_claim":"current methods exhibit consistently poor performance, revealing significant limitations in their robustness and generalization to unseen semantic content.","one_line_summary":"ImageAttributionBench is a benchmark dataset demonstrating that state-of-the-art image attribution methods lack robustness to image degradation and fail to generalize to semantically disjoint domains.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the two evaluation settings (training on balanced split then testing on degraded images, and training/testing on semantically disjoint splits) adequately simulate real-world attribution scenarios.","pith_extraction_headline":"Existing image attribution methods fail to generalize to unseen semantics and degraded images."},"references":{"count":88,"sample":[{"doi":"","year":2023,"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","ref_index":1,"cited_arxiv_id":"2303.08774","is_internal_anchor":true},{"doi":"","year":null,"title":"Ai dungeon: Ai-powered text adventure game.https://aidungeon.com/","work_id":"429692d4-aa98-4d4d-b5f0-dd99b2512b78","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond","work_id":"cbc2bb21-b6bb-46c0-80bf-107e195ffe10","ref_index":3,"cited_arxiv_id":"2308.12966","is_internal_anchor":true},{"doi":"","year":2023,"title":"Improving image generation with better captions.Computer Science","work_id":"fb7509a6-ece6-4ea3-b583-1ec884016dc8","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"FLUX.2 [klein]: Towards interactive visual intelligence, January 2026","work_id":"273a9cab-6f3b-46e3-9570-c9522c614b4f","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":88,"snapshot_sha256":"e027424e7a41151f68bcdc70a32a7d098e17c5b5d85d2f8d23bcaa49d07f0185","internal_anchors":13},"formal_canon":{"evidence_count":2,"snapshot_sha256":"26c9ffacf0b216bf31bc763d4ffeac1de983fbf4b7251a8ac2b6341a6e16fb6b"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}