Research map / Vision-language and multimodal / Multimodal LLMs
REVEAL: Reasoning-Enhanced Forensic Evidence Analysis for Explainable AI-Generated Image Detection
Huangsen Cao, Qin Mei, Zhiheng Li, Yuxi Li, Zhan Meng, Ying Zhang, Chen Li, Zhimeng Zhang, Xin Ding, Yongwei Wang, Jing LYU, Fei Wu
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 |
|---|---|---|---|
| add R-GRPO reward-based RL on top of CoE tuning (instead of vanilla GRPO) (ablation) | REVEAL · REVEAL-Bench | accuracy: 91.56 -> 95.31 | Improved |
| add vanilla GRPO after CoE tuning (ablation) | REVEAL · REVEAL-Bench | accuracy: 85.73 -> 91.56 | Improved |
| add chain-of-evidence tuning instead of no reasoning supervision (ablation) | REVEAL · REVEAL-Bench | accuracy: 61.21 -> 85.73 | Improved |
| use evidence-first CoE format instead of answer-first reasoning format (ablation) | REVEAL · REVEAL-Bench | accuracy: 82.39 -> 85.73 | Improved |
| add R-GRPO to Phi-3.5 after CoE tuning (ablation) | Phi-3.5 · REVEAL-Bench | accuracy: 83.75 -> 87.19 | Improved |
| add R-GRPO to Qwen2.5-VL-7B after CoE tuning (ablation) | Qwen2.5-VL-7B · REVEAL-Bench | accuracy: 85.73 -> 92.19 | Improved |
| add R-GRPO to llava-v1.5-7b after CoE tuning (ablation) | llava-v1.5-7b · REVEAL-Bench | accuracy: 91.56 -> 92.81 | Improved |
| scale backbone from LLaVA-7B to LLaVA-13B (with R-GRPO) (ablation) | REVEAL · REVEAL-Bench | accuracy: 92.81 -> 95.31 | Improved |
| use REVEAL's LLM reasoning over expert outputs instead of majority voting of the eight experts (ablation) | REVEAL · REVEAL-Bench | accuracy: 78.35 -> 95.31 | Improved |
| use REVEAL instead of decision tree over expert outputs (ablation) | REVEAL · REVEAL-Bench | accuracy: 74.75 -> 95.31 | Improved |
| add expert guidance to CoE annotation instead of direct LLM annotation (ablation) | REVEAL-Bench · annotation quality | label errors and explanation correctness | Improved |
| add the Pixel Noise expert to the annotation pipeline (ablation) | REVEAL-Bench · REVEAL-Bench | accuracy | Improved |
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.
- AIGI-Holmes — AIGI-Holmes: Towards Explainable and Generalizable AI-Generated Image Detection via Multimodal Large Language Models (ICCV 2025) · better on 3 of 3 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
- Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated Images (CVPR 2026 · similarity 0.88)
- A Bias-Free Training Paradigm for More General AI-generated Image Detection (CVPR 2025 · similarity 0.87)
- Community Forensics: Using Thousands of Generators to Train Fake Image Detectors (CVPR 2025 · similarity 0.87)
- BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation (NeurIPS 2026 · similarity 0.87)
- Seeing What Matters: Generalizable AI-generated Video Detection with Forensic-Oriented Augmentation (NeurIPS 2025 · similarity 0.86)
- THEMIS: Towards Holistic Evaluation of MLLMs for Scientific Paper Fraud Forensics (ICLR 2026 · similarity 0.86)
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.
