Research map / Vision-language and multimodal / CLIP and zero-shot
RefChess: Training-Free Contextual Search for Zero-Shot Referring Image Segmentation
Shiyan Tong, Jinxia Zhang, Zhiyuan Wang, Hao Tian, YingYing Wang, Kanjian Zhang, Haikun Wei
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 |
|---|---|---|---|
| replace greedy proposal selection with MCTS contextual search (ablation) | RefChess · RefCOCO val | oIoU: 29.95 -> 48.47 | Improved |
| replace greedy with MCTS (ablation) | RefChess · RefCOCO+ val | oIoU: 35.83 -> 41.20 | Improved |
| replace greedy with MCTS (ablation) | RefChess · RefCOCOg val | oIoU: 34.49 -> 42.85 | Improved |
| increase simulations from 256 to 4096 (ablation) | RefChess · RefCOCO val | mIoU: 53.84 -> 55.18 | Improved |
| increase simulations from 256 to 1024 (ablation) | RefChess · RefCOCO val | mIoU: 53.84 -> 53.85 | No clear effect |
| increase rollout depth from 5 to default (ablation) | RefChess · RefCOCO val | oIoU: 46.84 -> 48.47 | Improved |
| use hybrid GEM+FGCLIP heatmaps instead of GEM only (ablation) | RefChess · RefCOCO val | oIoU: 47.00 -> 48.47 | Improved |
| use hybrid heatmaps instead of FGCLIP only (ablation) | RefChess · RefCOCOg val | mIoU: 50.64 -> 50.63 | No clear effect |
| add size bias (ablation) | RefChess · RefCOCO+ val | oIoU: 38.56 -> 41.20 | Improved |
| add heatmap reward term (ablation) | RefChess · RefCOCO val | oIoU: 46.16 -> 48.47 | Improved |
| add full-expression matching term (ablation) | RefChess · RefCOCO val | oIoU: 45.99 -> 48.47 | Improved |
| add object-centric term (ablation) | RefChess · RefCOCO val | oIoU: 48.45 -> 48.47 | No clear effect |
Closest papers by meaning
- IteRPrimE: Zero-shot Referring Image Segmentation with Iterative Grad-CAM Refinement and Primary Word Emphasis (AAAI 2025 · similarity 0.91)
- Hybrid Global-Local Representation with Augmented Spatial Guidance for Zero-Shot Referring Image Segmentation (CVPR 2025 · similarity 0.90)
- Prompt-Driven Referring Image Segmentation with Instance Contrasting (CVPR 2024 · similarity 0.90)
- Early Semantic Grounding in Image Editing Models for Zero-Shot Referring Image Segmentation (NeurIPS 2026 · similarity 0.90)
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.90)
- Curriculum Point Prompting for Weakly-Supervised Referring Image Segmentation (CVPR 2024 · similarity 0.89)
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.
