Research map / Vision-language and multimodal / Multimodal LLMs
CORE-V: Chain-Of-thought REasoning for Image Editing with Visual Interaction
Tianyang Yan, Peisen Zhao, Guanghao Zheng, Mingxing Xu, Zhibo Zhang, Wenrui Dai, Junni Zou, Hongkai Xiong, Xiaopeng Zhang, Qi Tian
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 |
|---|---|---|---|
| use pretrained MLLM encoder instead of CLIP (ablation) | CORE-V · Reason-Edit | PSNR: 21.307 -> 23.257 | Improved |
| add textual CoT reasoning (ablation) | CORE-V with MLLM encoder · Reason-Edit | PSNR: 23.257 -> 23.544 | Improved |
| add visual imagination (reference images) (ablation) | CORE-V · Reason-Edit | PSNR: 23.544 -> 25.012 | Improved |
| add spatial control (mask, depth, edge) (ablation) | CORE-V · Reason-Edit | PSNR: 25.012 -> 24.567 | Mixed |
| add spatial control (mask, depth, edge) (ablation) | CORE-V · Reason-Edit | SSIM: 0.858 -> 0.880 | Improved |
| add spatial control (mask, depth, edge) (ablation) | CORE-V · Reason-Edit | CLIP: 22.46 -> 22.89 | Improved |
| raise lambda_edit from 2.5 to 3.0 (ablation) | CORE-V · Emu-Edit | CLIPout: 0.313 -> 0.309 | Hurt |
| raise lambda_edit from 2.5 to 4.0 (ablation) | CORE-V · Emu-Edit | CLIPout: 0.313 -> 0.297 | Hurt |
| reduce lambda_ref from 1.0 to 0.6 (ablation) | CORE-V · Emu-Edit | DINO: 0.885 -> 0.865 | Hurt |
| reduce lambda_ref from 1.0 to 0.8 (ablation) | CORE-V · Emu-Edit | DINO: 0.885 -> 0.880 | Hurt |
| CORE-V visual CoT vs BAGEL | CORE-V · Reason-Edit (reasoning) | PSNR: 27.331 -> 28.356 | Improved |
| CORE-V vs BAGEL at context level | CORE-V · CORE-Bench (context level) | CLIPDelta x100: 7.565 -> 7.901 | Improved |
Closest papers by meaning
- Reasoning to Edit: Hypothetical Instruction-Based Image Editing with Visual Reasoning (ICML 2026 · similarity 0.91)
- Envisioning Beyond the Pixels: Benchmarking Reasoning-Informed Visual Editing (NeurIPS 2025 · similarity 0.90)
- KRIS-Bench: Benchmarking Next-Level Intelligent Image Editing Models (NeurIPS 2025 · similarity 0.90)
- ReasonEdit: Towards Reasoning-Enhanced Image Editing Models (CVPR 2026 · similarity 0.89)
- EdiVal-Agent: An Object-Centric Framework for Automated, Fine-Grained Evaluation of Multi-Turn Editing (ICLR 2026 · similarity 0.88)
- Text-Vision Co-Instructed Image Editing (NeurIPS 2026 · similarity 0.88)
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.
