Research map / Detection and segmentation / Open-set and few-shot detection
Contrastive Grouping With Transformer for Referring Image Segmentation
Jiajin Tang, Ge Zheng, Cheng Shi, Sibei Yang
This paper tried a common change: Add a contrastive loss
Add grouping with contrastive loss to assign tokens to distinct regions · RefCOCO val · oIoU: 66.46 -> 70.81 Improved
What 1,317 accepted papers report for this change
- 1,253 improved
- 22 no clear effect
- 16 mixed
- 26 hurt
- Beyond Looking Up, Try Looking Around: Harmonizing Global Structure and Local Consistency in Optimal Transport for Short Text Clustering ICML 2026 · add contrastive learning · SearchSnippets · ACC: 56.65 -> 80.12Improved
- SleepLM: Natural-Language Intelligence for Human Sleep ICML 2026 · add contrastive objective to captioning (CoCa vs Cap) · sleep event IoU · IoU: 29.2 -> 27.4Hurt
- LIMSSR: LLM-Driven Sequence-to-Score Reasoning under Training-Time Incomplete Multimodal Observations ICML 2026 · add contrastive loss L_con · FS1000 average · Spearman correlation: 0.759 -> 0.789Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 CGFormer rather than LAVT | Referring image segmentation · RefCOCO val | oIoU: 72.73 -> 74.75 | Improved |
| Use CGFormer rather than LAVT | Referring image segmentation · RefCOCO+ val | oIoU: 62.14 -> 64.54 | Improved |
| Use CGFormer rather than LAVT | Referring image segmentation · G-Ref val-U | oIoU: 61.24 -> 64.68 | Improved |
| Use CGFormer rather than CEFNet | Referring image segmentation · ReferIt test | oIoU: 66.70 -> 73.36 | Improved |
| Add one linguistic-conditioned query token (ablation) | RefCOCO segmentation baseline · RefCOCO val | oIoU: 65.70 -> 66.39 | Improved |
| Use multiple query tokens instead of one (ablation) | RefCOCO segmentation model · RefCOCO val | oIoU: 66.39 -> 66.46 | No clear effect |
| Add grouping with contrastive loss to assign tokens to distinct regions (ablation) | RefCOCO segmentation model with multiple tokens · RefCOCO val | oIoU: 66.46 -> 70.81 | Improved |
| Add learnable Gumbel-Softmax hard assignment (ablation) | RefCOCO model with grouping · RefCOCO val | oIoU: 70.81 -> 72.44 | Improved |
| Add parallel multi-scale decoding (ablation) | RefCOCO model with hard assignment · RefCOCO val | oIoU: 72.44 -> 73.28 | Improved |
| Use Consecutive Decoder instead of parallel multi-scale querying (ablation) | RefCOCO multi-scale model · RefCOCO val | oIoU: 73.28 -> 74.75 | Improved |
| Use CGFormer rather than VLT with the same backbone and text encoder | RefCOCO segmentation · RefCOCO val | oIoU: 70.89 -> 74.75 | Improved |
| Replace cosine similarity with dot product (ablation) | CGFormer grouping · RefCOCO val | oIoU: 74.75 -> 73.37 | Hurt |
Closest papers by meaning
- Language-Aware Vision Transformer for Referring Segmentation (TPAMI 2024 · similarity 0.90)
- LQMFormer: Language-aware Query Mask Transformer for Referring Image Segmentation (CVPR 2024 · similarity 0.90)
- Vision-Language Transformer and Query Generation for Referring Segmentation (ICCV 2021 · similarity 0.89)
- VLT: Vision-Language Transformer and Query Generation for Referring Segmentation (TPAMI 2022 · similarity 0.89)
- See-Through-Text Grouping for Referring Image Segmentation (ICCV 2019 · similarity 0.88)
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.87)
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.
