Research map / Detection and segmentation / Open-set and few-shot detection
Toward Robust In-Context Segmentation via Concept Guidance
Zhigang Chen, Xiawu Zheng, Rongrong Ji
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 multiple concept candidates alone to single-concept prompting (ablation) | CG-ICS · COCO-20i (1-shot, reference-selection robustness) | mIoU: 67.6 -> 66.5 | No clear effect |
| add RF and QM concept scoring to candidates (ablation) | CG-ICS · COCO-20i (1-shot, reference-selection robustness) | mIoU: 66.5 -> 70.1 | Improved |
| add RF and QM concept scoring to candidates (ablation) | CG-ICS · COCO-20i (1-shot, reference-selection robustness) | CV: 23.0% -> 16.1% | Improved |
| add tree-search selection (ablation) | CG-ICS · COCO-20i (1-shot, reference-selection robustness) | CV: 16.1% -> 14.7% | Improved |
| add tree-search selection (ablation) | CG-ICS · COCO-20i (1-shot, reference-selection robustness) | mIoU: 70.1 -> 71.2 | Improved |
| add visual exemplar branch (ablation) | CG-ICS · COCO-20i (1-shot, reference-selection robustness) | mIoU: 71.2 -> 72.1 | Improved |
| add visual exemplar branch (ablation) | CG-ICS · COCO-20i (1-shot, reference-selection robustness) | CV: 14.7% -> 12.9% | Improved |
| increase number of search nodes N from 1 to 4 (ablation) | CG-ICS · COCO-20i (reference-selection robustness) | mIoU: 67.6 -> 71.9 | Improved |
| increase number of search nodes N from 1 to 4 (ablation) | CG-ICS · COCO-20i (reference-selection robustness) | Std: 15.1 -> 10.5 | Improved |
| increase N beyond 5 (ablation) | CG-ICS · COCO-20i (reference-selection robustness) | mIoU: 72.10 -> 72.4 | No clear effect |
| increase search rounds K beyond 3 (ablation) | CG-ICS · COCO-20i (reference-selection robustness) | mIoU | No clear effect |
| use concept-guided CG-ICS instead of GF-SAM | CG-ICS vs GF-SAM · Pascal-5i (1-shot, standard ICS) | mIoU: 72.1 -> 89.3 | 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.
- SANSA — SANSA: Unleashing the Hidden Semantics in SAM2 for Few-Shot Segmentation (NeurIPS 2025) · better on 2 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
- UNICL-SAM: Uncertainty-Driven In-Context Segmentation with Part Prototype Discovery (CVPR 2025 · similarity 0.88)
- From Pixels to Concepts: Do Segmentation Models Understand What They Segment? (NeurIPS 2026 · similarity 0.86)
- Seg-ReSearch: Segmentation with Interleaved Reasoning and External Search (ICML 2026 · similarity 0.86)
- Hybrid Global-Local Representation with Augmented Spatial Guidance for Zero-Shot Referring Image Segmentation (CVPR 2025 · similarity 0.86)
- Uncertainty-aware Fine-tuning of Segmentation Foundation Models (NeurIPS 2024 · similarity 0.86)
- SegCompass: Exploring Interpretable Alignment with Sparse Autoencoders for Enhanced Reasoning Segmentation (CVPR 2026 · similarity 0.85)
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.
