Research map / Vision-language and multimodal / Multimodal LLMs
Code-in-the-Loop Forensics: Agentic Tool Use for Image Forgery Detection
Fanrui Zhang, Qiang Zhang, Sizhuo Zhou, Jianwen Sun, Chuanhao Li, Jiaxin Ai, Yukang Feng, Yujie Zhang, Wenjie Li, Zizhen Li, Yifan Chang, Jiawei Liu and 1 more
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 Cold-Start SFT stage before RFT (ablation) | ForenAgent · FABench | accuracy (%): 78.7 -> 89.5 | Improved |
| add RFT with GRPO after cold start (ablation) | ForenAgent · FABench | accuracy (%): 81.4 -> 89.5 | Improved |
| add tool usage reward (ablation) | ForenAgent RFT · FABench | accuracy (%): 83.8 -> 89.5 | Improved |
| add tool logic reward (ablation) | ForenAgent RFT · FABench | accuracy (%): 84.4 -> 89.5 | Improved |
| add global forensic priority reward (ablation) | ForenAgent RFT · FABench | accuracy (%): 86.4 -> 89.5 | Improved |
| add crop sensitivity reward (ablation) | ForenAgent RFT · FABench | accuracy (%): 87.8 -> 89.5 | Improved |
| add reasoning coherence reward (ablation) | ForenAgent RFT · FABench | accuracy (%): 88.0 -> 89.5 | Improved |
| use Qwen2.5-VL-7B instead of InternVL3-8B backbone (ablation) | ForenAgent · FABench | accuracy (%): 79.8 -> 89.5 | Improved |
| use Qwen2.5-VL-7B instead of LLaVA-7B backbone (ablation) | ForenAgent · FABench | accuracy (%): 75.8 -> 89.5 | Improved |
| equip zero-shot MLLM with forensic toolbox | GPT-4.1 · FABench | accuracy (%): 51.4 -> 64.7 | Improved |
| equip zero-shot MLLM with forensic toolbox | Qwen2.5-VL-72B · FABench | accuracy (%): 50.0 -> 57.4 | Improved |
| ForenAgent vs trained Qwen2.5-VL-7B without tools | Qwen2.5-VL-7B · SIDA-Test | accuracy (%): 72.9 -> 83.1 | Improved |
Closest papers by meaning
- A Hitchhiker's Guide to Fine-Grained Face Forgery Detection Using Common Sense Reasoning (NeurIPS 2024 · similarity 0.87)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.86)
- ForgerySleuth: Empowering Multimodal Large Language Models for Image Manipulation Detection (NeurIPS 2025 · similarity 0.86)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.86)
- Forensics-Bench: A Comprehensive Forgery Detection Benchmark Suite for Large Vision Language Models (CVPR 2025 · similarity 0.86)
- Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated Images (CVPR 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.
