Research map / Large language models / AI and society
Learning Perceptual Inference by Contrasting
Chi Zhang, Baoxiong Jia, Feng Gao, Yixin Zhu, HongJing Lu, Song-Chun Zhu
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 |
|---|---|---|---|
| remove the perceptual inference branch (ablation) | CoPINet · RAVEN test | accuracy: 91.42% -> 90.04% | Hurt |
| replace contrast loss with cross-entropy loss (ablation) | CoPINet · RAVEN test | accuracy: 91.42% -> 86.16% | Hurt |
| remove the contrast module and use cross-entropy loss (ablation) | CoPINet · RAVEN test | accuracy: 91.42% -> 20.75% | Hurt |
| remove the perceptual inference branch (ablation) | CoPINet · PGM neutral test | accuracy: 56.37% -> 54.19% | Hurt |
| replace contrast loss with cross-entropy loss (ablation) | CoPINet · PGM neutral test | accuracy: 56.37% -> 51.04% | Hurt |
| remove the contrast module and use cross-entropy loss (ablation) | CoPINet · PGM neutral test | accuracy: 56.37% -> 42.10% | Hurt |
| add auxiliary loss (ablation) | WReN-NoTag · PGM neutral test | accuracy: 39.25% -> 49.10% | Improved |
| add positional tagging (ablation) | WReN-NoAux · PGM neutral test | accuracy: 39.25% -> 62.45% | Improved |
| add positional tagging and auxiliary loss (ablation) | WReN · PGM neutral test | accuracy: 39.25% -> 77.94% | Improved |
| add positional tagging and auxiliary loss (ablation) | WReN · RAVEN test | accuracy: 15.07% -> 33.97% | Improved |
| add auxiliary loss (ablation) | WReN-Tag · RAVEN test | accuracy: 17.94% -> 33.97% | Improved |
Closest papers by meaning
- Imagination Helps Visual Reasoning, But Not Yet in Latent Space (ICML 2026 · similarity 0.87)
- Thinking with Imagination: Agentic Visual Spatial Reasoning with World Simulators (NeurIPS 2026 · similarity 0.86)
- A Framework for Inference Inspired by Human Memory Mechanisms (ICLR 2024 · similarity 0.86)
- Do multimodal models imagine electric sheep? (NeurIPS 2026 · similarity 0.86)
- Mind the GAP: Glimpse-based Active Perception improves generalization and sample efficiency of visual reasoning (ICLR 2025 · similarity 0.86)
- Linear Mechanisms for Spatiotemporal Reasoning in Vision Language Models (ICLR 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.
