Research map / Vision-language and multimodal / Multimodal LLMs
DiffTell: A High-Quality Dataset for Describing Image Manipulation Changes
Zonglin Di, Jing Shi, Yifei Fan, Hao Tan, Alexander Black, John Collomosse, Yang Liu
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 DiffTell to IER fine-tuning | Qwen2-VL-7B · IER test | CIDEr: 64.61 -> 67.71 | Improved |
| add DiffTell to IER fine-tuning | Fuyu-8B · IER test | CIDEr: 23.67 -> 41.05 | Improved |
| add DiffTell to IER fine-tuning | Idefics3-8B · IER test | CIDEr: 53.06 -> 64.35 | Improved |
| add DiffTell to IER fine-tuning | CLIP4IDC · IER test | CIDEr: 22.52 -> 28.14 | Improved |
| add DiffTell to IER fine-tuning | Qwen2-VL-7B · PSBattle (zero-shot) | CIDEr: 14.29 -> 16.66 | Improved |
| add DiffTell to IER fine-tuning | Fuyu-8B · PSBattle (zero-shot) | CIDEr: 4.05 -> 4.19 | No clear effect |
| add MagicBrush subset to IER training (ablation) | Qwen2-VL-7B · IER test | CIDEr: 64.61 -> 59.31 | Hurt |
| add COCO subset to IER training (ablation) | Qwen2-VL-7B · IER test | CIDEr: 64.61 -> 59.92 | Hurt |
| add OCR (MARIO-10M) subset to IER training (ablation) | Qwen2-VL-7B · IER test | CIDEr: 64.61 -> 62.00 | Hurt |
| add InstructPix2Pix subset to IER training (ablation) | Qwen2-VL-7B · IER test | CIDEr: 64.61 -> 66.23 | Improved |
| add GIER subset to IER training (ablation) | Qwen2-VL-7B · IER test | CIDEr: 64.61 -> 66.97 | Improved |
| add 10K randomly selected unfiltered MARIO-10M samples (ablation) | Qwen2-VL-7B · IER test | CIDEr: 64.61 -> 64.31 | No clear effect |
Closest papers by meaning
- Img-Diff: Contrastive Data Synthesis for Multimodal Large Language Models (CVPR 2025 · similarity 0.89)
- Describing Differences in Image Sets with Natural Language (CVPR 2024 · similarity 0.89)
- An Item Is Worth a Prompt: Versatile Image Editing with Disentangled Control (AAAI 2025 · similarity 0.87)
- VisMin: Visual Minimal-Change Understanding (NeurIPS 2024 · similarity 0.87)
- VIXEN: Visual Text Comparison Network for Image Difference Captioning (AAAI 2024 · similarity 0.87)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · 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.
