Research map / Vision-language and multimodal / Human and embodied modeling
Few-shot Visual Reasoning with Meta-Analogical Contrastive Learning
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 analogy learning to CoPINet | CoPINet · RAVEN test | accuracy (%): 91.42 -> 93.06 | Helped |
| Add analogy learning to LSTM | LSTM · RAVEN test | accuracy (%): 13.07 -> 20.70 | Helped |
| Add analogy learning to CNN | CNN · RAVEN test | accuracy (%): 36.97 -> 50.37 | Helped |
| Use generative analogy (ablation) | CoPINet with analogy learning · RAVEN test with 35–651 training samples | accuracy (%): 11.07 -> 18.99; 13.49 -> 27.27; 21.01 -> 40.11; 29.47 -> 51.96; 40.34 -> 61.95 | Helped |
| Use inference analogy (ablation) | CoPINet with analogy learning · RAVEN test, 14–651 training samples | accuracy (%): 6.26 -> 7.48; 11.07 -> 14.10; 13.49 -> 21.31; 21.01 -> 27.27; 29.47 -> 33.77; 40.34 -> 48.30 | Helped |
| Use variational analogy (ablation) | CoPINet with analogy learning · RAVEN test, 14–651 training samples | accuracy (%): 6.26 -> 5.60; 11.07 -> 12.35; 13.49 -> 21.30; 21.01 -> 25.19; 29.47 -> 35.54; 40.34 -> 48.81 | Mixed |
| Add analogy learning without a kernel process (ablation) | CoPINet · RAVEN test with 14–651 training samples | accuracy (%): 6.26 -> 7.68; 11.07 -> 13.71; 13.49 -> 20.64; 21.01 -> 27.32; 29.47 -> 33.12; 40.34 -> 48.09 | Helped |
| Use generative analogy (ablation) | CoPINet · RAVEN unseen shapes, center configuration, minibatch 2 | accuracy (%): 48.75 -> 73.75 | Helped |
| Use generative analogy (ablation) | CoPINet · RAVEN unseen shapes, center configuration, minibatch 32 | accuracy (%): 25.00 -> 43.75 | Helped |
| Use minibatch size 2 rather than 32 (ablation) | CoPINet with analogy variants · RAVEN unseen shapes, center configuration | accuracy (%): 48.75 -> 25.00; 58.75 -> 51.25; 61.25 -> 45.00; 56.25 -> 43.75; 73.75 -> 43.75 | Mixed |
| Add meta-contrastive learning to pointwise analogy variants (ablation) | CoPINet analogy variants · RAVEN unseen shapes, center configuration | accuracy (%): 58.75 -> 61.25; 51.25 -> 50.00; 61.25 -> 62.50; 45.00 -> 55.00; 56.25 -> 62.50; 43.75 -> 53.75; 73.75 -> 63.75; 43.75 -> | Mixed |
| Add meta-contrastive learning to the CoPINet baseline (ablation) | CoPINet · RAVEN unseen shapes, center configuration | accuracy (%): 48.75 -> 51.25; 25.00 -> 48.75 | Helped |
Closest papers by meaning
- MiCo: Multi-image Contrast for Reinforcement Visual Reasoning (NeurIPS 2025 · similarity 0.88)
- VOILA: Evaluation of MLLMs For Perceptual Understanding and Analogical Reasoning (ICLR 2025 · similarity 0.88)
- Through the Lens of Contrast: Self-Improving Visual Reasoning in VLMs (ICLR 2026 · similarity 0.87)
- AStar: Boosting Multimodal Reasoning with Automated Structured Thinking (AAAI 2026 · similarity 0.87)
- SLVR: Structured Latent Visual Reasoning via Human-like Reasoning Flows (NeurIPS 2026 · similarity 0.86)
- Mimic Human Cognition, Master Multi-Image Reasoning: A Meta-Action Framework for Enhanced Visual Understanding (CVPR 2026 · similarity 0.86)
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.