Research map / Label-efficient and robust learning / OOD and anomaly detection
Adaptive Contextual Perception: How To Generalize To New Backgrounds and Ambiguous Objects
Zhuofan Ying, Peter Hase, Mohit Bansal
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 |
|---|---|---|---|
| rotate foreground subspace into background subspace (reduce factorization) (ablation) | ResNet features · SceneObject | factorization (geometric): drop by 85.63\% | Hurt |
| same rotation intervention (ablation) | ResNet features · SceneObject | Background-Invariance accuracy: drop by 18.24\% | Hurt |
| same rotation intervention (ablation) | ResNet features · SceneObject | Object-Disambiguation accuracy: increases by 5.97\% | Improved |
| decrease foreground feature weighting by factor 64 (ablation) | linear probe on features · SceneObject | Background-Invariance accuracy: decreases by 0.63 points | Hurt |
| decrease foreground feature weighting by factor 64 (ablation) | linear probe on features · SceneObject | Object-Disambiguation accuracy: increases by 0.68 points | Improved |
| decrease background feature weighting by factor 64 (ablation) | linear probe on features · SceneObject | Background-Invariance accuracy: increases by 0.35 points | Improved |
| decrease background feature weighting by factor 64 (ablation) | linear probe on features · SceneObject | Object-Disambiguation accuracy: decreases by 0.92 points | Hurt |
| use proposed factorization/feature-weighting augmentation instead of ERM | ResNet · SceneObject | Object-Disambiguation accuracy: 71.93±1.27 -> 73.86±1.11 | Improved |
| use proposed augmentation instead of best baseline Fish | ResNet · SceneObject | Object-Disambiguation accuracy: 71.30±0.74 -> 73.86±1.11 | Improved |
| use proposed augmentation instead of ERM | ResNet · ColorObject | Background-Invariance accuracy: 29.11±3.39 -> 42.97±3.02 | Improved |
| use proposed augmentation instead of ERM | ResNet · ColorObject | Object-Disambiguation accuracy: 95.24±0.83 -> 92.02±1.06 | Hurt |
| use proposed augmentation instead of ERM | ResNet · ColorObject | factorization: 71.27±1.14 -> 76.93±1.33 | Improved |
Closest papers by meaning
- Characterizing Universal Object Representations Across Vision Models (NeurIPS 2026 · similarity 0.87)
- Overcoming the Pitfalls of Vision-Language Model Finetuning for OOD Generalization (ICLR 2024 · similarity 0.87)
- Benchmarking Out-of-Distribution Generalization Capabilities of DNN-based Encoding Models for the Ventral Visual Cortex. (NeurIPS 2024 · similarity 0.87)
- Flexible Context-Driven Sensory Processing in Dynamical Vision Models (NeurIPS 2024 · similarity 0.86)
- Does Data Scaling Lead to Visual Compositional Generalization? (ICML 2025 · similarity 0.86)
- In Search of Forgotten Domain Generalization (ICLR 2025 · 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.
