Research map / Graph learning / Knowledge and temporal graphs
Iterative Visual Reasoning Beyond Convolutions
Xinlei Chen, Li-Jia Li, Li Fei-Fei, Abhinav Gupta
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 local spatial-memory reasoning | ResNet-50 region classifier · ADE test-1k | per-class AP: 40.1 -> 47.9 | Improved |
| Combine local and global iterative reasoning | ResNet-50 region classifier · ADE test-1k | per-class AP: 40.1 -> 48.5 | Improved |
| Add local and global reasoning | ResNet-50 region classifier · Visual Genome test | per-class AP: 16.9 -> 19.1 | Improved |
| Add local and global reasoning | ResNet-50 region classifier · COCO minival | per-class AP: 83.7 -> 87.4 | Improved |
| Add hard-example re-weighting (ablation) | Local reasoning module · ADE test-1k | per-class AP: 46.7 -> 47.9 | Improved |
| Add local context aggregator C (ablation) | Local reasoning module · ADE test-1k | per-class AP: 46.1 -> 47.9 | Improved |
| Add spatial memory S (ablation) | Local reasoning module · ADE test-1k | per-class AP: 42.1 -> 47.9 | Improved |
| Add spatial region-to-region graph edges (ablation) | Global reasoning module · ADE test-1k | per-class AP: 41.5 -> 44.5 | Improved |
| Add semantic class-to-class graph edges (ablation) | Global reasoning module · ADE test-1k | per-class AP: 43.9 -> 44.5 | Improved |
| Add graph reasoning module R (ablation) | Global reasoning module · ADE test-1k | per-class AP: 41.5 -> 44.5 | Improved |
| Add global memory M alongside graph reasoning (ablation) | Global reasoning module · ADE test-1k | per-class AP: 41.0 -> 41.5 | No clear effect |
| Add hard-example re-weighting (ablation) | Full local-global pipeline · ADE test-1k | per-class AP: 47.3 -> 48.5 | Improved |
Closest papers by meaning
- Learning to Reason Iteratively and Parallelly for Complex Visual Reasoning Scenarios (NeurIPS 2024 · similarity 0.89)
- ATLAS: Agentic or Latent Visual Reasoning? One Word is Enough for Both (NeurIPS 2026 · similarity 0.88)
- Visual Graph Arena: Evaluating Visual Conceptualization of Vision and Multimodal Large Language Models (ICML 2025 · similarity 0.88)
- No Labels, No Problem: Training Visual Reasoners with Multimodal Verifiers (ICLR 2026 · similarity 0.88)
- Multi-step Visual Reasoning with Visual Tokens Scaling and Verification (NeurIPS 2025 · similarity 0.87)
- SLVR: Structured Latent Visual Reasoning via Human-like Reasoning Flows (NeurIPS 2026 · 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.
