Research map / Image generation / Image restoration
See and Fix the Flaws: Enabling VLMs and Diffusion Models to Comprehend Visual Artifacts via Agentic Data Synthesis
Jaehyun Park, Minyoung Ahn, Minkyu Kim, Jonghyun Lee, Jae-Gil Lee, Dongmin Park
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune on ArtiAgent data | InternVL3.5-8B · ArtiBench detection | accuracy: 0.498 -> 0.630 | Improved |
| fine-tune on ArtiAgent data | Qwen2.5-VL-7B · ArtiBench detection | accuracy: 0.501 -> 0.627 | Improved |
| fine-tune on ArtiAgent data | Qwen2.5-VL-7B · ArtiBench localization | mIoU: 0.010 -> 0.111 | Improved |
| fine-tune on ArtiAgent data | Qwen2.5-VL-7B · ArtiBench explanation | ROUGE: 0.117 -> 0.233 | Improved |
| ArtiAgent-tuned open VLM vs GPT-5 on RichHF localization | Qwen2.5-VL-7B + ArtiAgent · RichHF localization | mIoU: 0.126 (GPT-5) vs 0.119 | Hurt |
| fine-tune on ArtiAgent data | InternVL3.5-8B · RichHF localization | mIoU: 0.013 -> 0.100 | Improved |
| increase ArtiAgent training data from 1K to 100K (ablation) | Qwen2.5-VL-7B · detection, localization, explanation | average across benchmarks | Improved |
| replace human SynthScars half by ArtiAgent samples (1K) (ablation) | Qwen2.5-VL-7B · ArtiBench detection | accuracy: 0.555 -> 0.555 | No clear effect |
| replace human SynthScars half by ArtiAgent samples (1K) (ablation) | Qwen2.5-VL-7B · ArtiBench localization | mIoU: 0.094 -> 0.074 | Hurt |
| replace human SynthScars half by ArtiAgent samples (1K) (ablation) | Qwen2.5-VL-7B · ArtiBench explanation | ROUGE: 0.156 -> 0.222 | Improved |
| test-time search with ArtiAgent reward model | FLUX-schnell · MS-COCO prompts | delta reward: 0 -> 0.23±0.08 | Improved |
| use DiffDoctor trained on human labels | DiffDoctor · LOKI vs ArtiBench localization | mIoU: 0.175 -> 0.081 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LEGION — LEGION: Learning to Ground and Explain for Synthetic Image Detection (ICCV 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SALART-VQA: Diagnosing Whether VLMs Understand Salient Artifacts in Generated Images (NeurIPS 2026 · similarity 0.89)
- Bifrost-1: Bridging Multimodal LLMs and Diffusion Models with Patch-level CLIP Latents (NeurIPS 2025 · similarity 0.85)
- Temporal Score Analysis for Understanding and Correcting Diffusion Artifacts (CVPR 2025 · similarity 0.85)
- Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models (CVPR 2025 · similarity 0.85)
- Self-Adaptive Reality-Guided Diffusion for Artifact-Free Super-Resolution (CVPR 2024 · similarity 0.85)
- I Think, Therefore I Diffuse: Enabling Multimodal In-Context Reasoning in Diffusion Models (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
