Research map / Vision-language and multimodal / Multimodal LLMs
CoReS: Orchestrating the Dance of Reasoning and Segmentation
Xiaoyi Bao, Siyang Sun, Shuailei Ma, Kecheng Zheng, Yuxin Guo, Guosheng Zhao, Yun Zheng, Xingang Wang
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 chain-of-segmenting guided by chain-of-reasoning to in-context input (ablation) | CoReS · ReasonSeg val | gIoU: 52.9 -> 59.4 | Improved |
| add token refinement to chain-of-segmenting (ablation) | CoReS · ReasonSeg val | gIoU: 58.4 -> 59.4 | Improved |
| add in-context input to dual-chain structure (ablation) | CoReS · ReasonSeg val | gIoU: 57.5 -> 59.4 | Improved |
| use chain-of-segmenting without chain-of-reasoning (ablation) | CoReS · ReasonSeg val | gIoU: 52.9 -> 54.0 | Improved |
| increase number of in-context examples beyond one (ablation) | CoReS · ReasonSeg val | gIoU: 59.4 -> 59.4 (2), 59.3 (4) | No clear effect |
| add a third logical level ([POS]) to the chain (ablation) | CoReS · ReasonSeg val | gIoU: 59.4 -> 59.7 | No clear effect |
| add text-only explicit CoT before segmentation (LISA) (ablation) | LISA · ReasonSeg val | gIoU: 52.9 -> 44.7 | Hurt |
| supervise intermediate chain mask with final mask (ablation) | CoReS · ReasonSeg val | gIoU: 59.4 -> 55.9 | Hurt |
| use query-specific prompts instead of random in-context examples (ablation) | CoReS · ReasonSeg val | gIoU: 59.4 -> 60.1 | Improved |
| replace LISA with CoReS on refCOCO+ | LLaVA-7B-v0 · refCOCO+ val | cIoU: 65.1 -> 65.1 | No clear effect |
| replace LISA with CoReS on refCOCOg (long queries) | LLaVA-7B-v0 · refCOCOg val(U) | cIoU: 67.9 -> 69.0 | Improved |
| increase weight of reasoning chain loss over segmentation loss (ablation) | CoReS · ReasonSeg val | gIoU: 59.4 -> 55.5 | Hurt |
Closest papers by meaning
- CoPRS: Learning Positional Prior from Chain-of-Thought for Reasoning Segmentation (ICLR 2026 · similarity 0.91)
- LISA: Reasoning Segmentation via Large Language Model (CVPR 2024 · similarity 0.90)
- DR$^2$Seg: Decomposed Two-Stage Rollouts for Efficient Reasoning Segmentation in Multimodal Large Language Models (ICML 2026 · similarity 0.89)
- Discriminative Perception via Anchored Description for Reasoning Segmentation (CVPR 2026 · similarity 0.89)
- SLVR: Structured Latent Visual Reasoning via Human-like Reasoning Flows (NeurIPS 2026 · similarity 0.89)
- Fast Reasoning Segmentation for Images and Videos (CVPR 2026 · 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.
