{"paper":{"title":"Degradation-Consistent Paired Training for Robust AI-Generated Image Detection","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Degradation-Consistent Paired Training raises AI-generated image detector accuracy on corrupted inputs by 9.1 percentage points with only a 0.9 percent drop on clean images.","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Xiaokun Yang, Yinghan Hou, Zongyou Yang","submitted_at":"2026-04-11T08:52:28Z","abstract_excerpt":"AI-generated image detectors suffer significant performance degradation under real-world image corruptions such as JPEG compression, Gaussian blur, and resolution downsampling. We observe that state-of-the-art methods, including B-Free, treat degradation robustness as a byproduct of data augmentation rather than an explicit training objective. In this work, we propose Degradation-Consistent Paired Training (DCPT), a simple yet effective training strategy that explicitly enforces robustness through paired consistency constraints. For each training image, we construct a clean view and a degraded"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on the Synthbuster benchmark (9 generators, 8 degradation conditions) demonstrate that DCPT improves the degraded-condition average accuracy by 9.1 percentage points compared to an identical baseline without paired training, while sacrificing only 0.9% clean accuracy. The improvement is most pronounced under JPEG compression (+15.7% to +17.9%).","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the specific degradations (JPEG compression, Gaussian blur, resolution downsampling) and consistency losses used during training will generalize to unseen real-world corruptions and that the Synthbuster benchmark sufficiently represents practical deployment conditions.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DCPT raises average accuracy of AI image detectors under real-world degradations by 9.1 points on the Synthbuster benchmark using paired consistency losses, with only 0.9% drop on clean images and no added parameters or inference cost.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Degradation-Consistent Paired Training raises AI-generated image detector accuracy on corrupted inputs by 9.1 percentage points with only a 0.9 percent drop on clean images.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"be26887965bba4e1e5219d36dd6307a869e58a329224a6a3e025290d17b2d58e"},"source":{"id":"2604.10102","kind":"arxiv","version":2},"verdict":{"id":"963d075e-8d30-4e3a-a957-30b6105a4db2","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T16:12:16.091897Z","strongest_claim":"Experiments on the Synthbuster benchmark (9 generators, 8 degradation conditions) demonstrate that DCPT improves the degraded-condition average accuracy by 9.1 percentage points compared to an identical baseline without paired training, while sacrificing only 0.9% clean accuracy. The improvement is most pronounced under JPEG compression (+15.7% to +17.9%).","one_line_summary":"DCPT raises average accuracy of AI image detectors under real-world degradations by 9.1 points on the Synthbuster benchmark using paired consistency losses, with only 0.9% drop on clean images and no added parameters or inference cost.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the specific degradations (JPEG compression, Gaussian blur, resolution downsampling) and consistency losses used during training will generalize to unseen real-world corruptions and that the Synthbuster benchmark sufficiently represents practical deployment conditions.","pith_extraction_headline":"Degradation-Consistent Paired Training raises AI-generated image detector accuracy on corrupted inputs by 9.1 percentage points with only a 0.9 percent drop on clean images."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.10102/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":2,"snapshot_sha256":"483efbc719a53b9aa393b9fe5dc5efbf20147727a685ea80e6634342d3f6b649"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}