Research map / Vision-language and multimodal / Human and embodied modeling
Visual Search Asymmetry: Deep Nets and Humans Share Similar Inherent Biases
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 eccentricity-dependent pooling instead of VGG16 max-pooling (ablation) | eccNET · Six visual-search asymmetry experiments | average Asymmetry Index | Helped |
| Combine top-down modulation across multiple layers instead of using only layer 18 to layer 17 (ablation) | eccNET · Six visual-search asymmetry experiments | average Asymmetry Index: 0.084 | Helped |
| Use a ResNet152 backbone instead of VGG16 (ablation) | eccNET · Six visual-search asymmetry experiments | experiments with positive asymmetry index: four out of the six | Hurt |
| Train from scratch on MNIST instead of ImageNet | eccNET · Lighting Direction (Experiment 2) | visual-search asymmetry | Hurt |
| Train from scratch on ImageNet rotated by 90 degrees | eccNET · Lighting Direction (Experiment 2) | polarity of asymmetry | Mixed |
| Train on ImageNet after a fisheye transform that increases curves and reduces straight lines | eccNET · Curvature (Experiment 1) | polarity of asymmetry | Mixed |
| Increase the proportion of straight lines in the training data | eccNET · Curvature (Experiment 1) | polarity of asymmetry | No clear effect |
| Train on Places365 instead of ImageNet | eccNET · Visual-search asymmetry experiments | polarity of asymmetry | Mixed |
| Train on rotated Places365 instead of ImageNet | eccNET · Visual-search asymmetry experiments | polarity of asymmetry | Mixed |
| Use eccNET rather than the baseline models | visual-search models · Six visual-search asymmetry experiments | agreement with human asymmetry: 5 out of 6 | Helped |
| Use target-dependent top-down features rather than bottom-up saliency alone | visual-search models · Six visual-search asymmetry experiments | agreement with human asymmetry | Helped |
| Use natural-image features from IVSN rather than chance, pixel matching, or GBVS | visual-search models · Six visual-search asymmetry experiments | Asymmetry Index | Helped |
Closest papers by meaning
- Gazing at Rewards: Eye Movements as a Lens into Human and AI Decision-Making in Hybrid Visual Foraging (CVPR 2025 · similarity 0.86)
- Beyond Accuracy: Tracking more like Human via Visual Search (NeurIPS 2024 · similarity 0.85)
- Directional Confusions Reveal Divergent Inductive Biases Through Rate-Distortion Geometry in Human and Machine Vision (NeurIPS 2026 · similarity 0.85)
- Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search (ICLR 2026 · similarity 0.85)
- Scaling Laws for Task-Optimized Models of the Primate Visual Ventral Stream (ICML 2025 · similarity 0.85)
- FlySearch: Exploring how vision-language models explore (NeurIPS 2025 · similarity 0.84)
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.