Research map / Vision-language and multimodal / Multimodal LLMs
Region-Based Global Reasoning Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 one ReGr module to Res50 | Res50 · CUB200 | Top-1 accuracy: 84.61% -> 86.19% | Helped |
| Add one ReGr module to Res50 | Res50 · UCF101 | Top-1 accuracy: 82.80% -> 83.98% | Helped |
| Remove refine operation (ablation) | Res50 + ReGr · CUB200 | Top-1 accuracy: 86.19% -> 85.88% | Hurt |
| Remove refine operation (ablation) | Res50 + ReGr · UCF101 | Top-1 accuracy: 83.98% -> 83.12% | Hurt |
| Use scale (4,4) instead of (1,1) pixel-level relations (ablation) | Res50 + ReGr · CUB200 | Top-1 accuracy: 85.40% -> 85.69% | Helped |
| Use scale (4,4) instead of (7,7) (ablation) | Res50 + ReGr · CUB200 | Top-1 accuracy: 85.55% -> 85.69% | Helped |
| Combine scales (4,4) and (7,7) (ablation) | Res50 + ReGr · CUB200 | Top-1 accuracy: 85.69% -> 86.19% | Helped |
| Use too many scales (ablation) | Res50 + ReGr · CUB200 | Top-1 accuracy: 86.19% -> 85.47% | Hurt |
| Combine scales (3,3) and (5,5) instead of single scale (ablation) | Res50 + ReGr · UCF101 | Top-1 accuracy: 83.79% -> 83.98% | Helped |
| Use too many scales (ablation) | Res50 + ReGr · UCF101 | Top-1 accuracy: 83.98% -> 83.95% | No clear effect |
| Use bilinear pooling instead of GCN or attention for relations (ablation) | Res50 + ReGr · CUB200 | Top-1 accuracy: 85.92% -> 86.19% | Helped |
| Use bilinear pooling instead of attention (ablation) | Res50 + ReGr · UCF101 | Top-1 accuracy: 82.81% -> 83.98% | Helped |
Closest papers by meaning
- RegionReasoner: Region-Grounded Multi-Round Visual Reasoning (ICLR 2026 · similarity 0.88)
- Recognize Any Regions (NeurIPS 2024 · similarity 0.87)
- Recognition through Reasoning: Reinforcing Image Geo-localization with Large Vision-Language Models (NeurIPS 2025 · similarity 0.85)
- VLM-R³: Region Recognition, Reasoning, and Refinement for Enhanced Multimodal Chain-of-Thought (NeurIPS 2025 · similarity 0.85)
- SpatialLadder: Progressive Training for Spatial Reasoning in Vision-Language Models (ICLR 2026 · similarity 0.85)
- GTR-Bench: Evaluating Geo-Temporal Reasoning in Vision-Language Models (ICLR 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.