Research map / Vision-language and multimodal / Multimodal LLMs
GAMR: A Guided Attention Model for (visual) Reasoning
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 |
|---|---|---|---|
| Use guided attention rather than self-attention (ablation) | GAMR · SVRT same-different tasks | relative test accuracy: 11.1% | Helped |
| Use guided attention and memory rather than no attention or memory (ablation) | GAMR · SVRT same-different tasks | relative test accuracy: 35.6% | Helped |
| Use guided attention rather than self-attention (ablation) | GAMR · SVRT spatial-relation tasks | relative test accuracy: 4.5% | Helped |
| Use guided attention and memory rather than no attention or memory (ablation) | GAMR · SVRT spatial-relation tasks | relative test accuracy: 10.4% | Helped |
| Add instance normalization (ablation) | GAMR · SVRT reasoning tasks | test accuracy | Helped |
| Add the controller output to the decision process (ablation) | GAMR · SVRT same-different tasks | test accuracy | Helped |
| Retain the relational vector in the decision process (ablation) | GAMR · SVRT | test accuracy | Helped |
| Retain the guided-attention vector when producing context vectors (ablation) | GAMR · SVRT | test accuracy | Helped |
| Use feature-channel gating rather than equal weighting (ablation) | GAMR · SVRT | test accuracy | Helped |
| Transfer learned routines across tasks with a shared abstract rule | GAMR · SVRT zero-shot task transfer | test accuracy | Helped |
| Increase test image scale from 1 to 1.5 without fine-tuning | GAMR · SVRT scale robustness | test accuracy: 94.77 -> 71.01; 97.34 -> 75.09 | Hurt |
| Add Gaussian white noise at 0 mean and .05 variance | GAMR · SVRT noise robustness | test accuracy | No clear effect |
Closest papers by meaning
- DeepEyes: Incentivizing "Thinking with Images" via Reinforcement Learning (ICLR 2026 · similarity 0.90)
- Mind the GAP: Glimpse-based Active Perception improves generalization and sample efficiency of visual reasoning (ICLR 2025 · similarity 0.89)
- Zebra-CoT: A Dataset for Interleaved Vision-Language Reasoning (ICLR 2026 · similarity 0.89)
- CoFFT: Chain of Foresight-Focus Thought for Visual Language Models (NeurIPS 2025 · similarity 0.89)
- Look on Demand: A Cognitive Scheduling Framework for Visual Evidence Acquisition in Multimodal Reasoning (ICML 2026 · similarity 0.89)
- From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models (ICML 2026 · similarity 0.88)
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.