Research map / Vision-language and multimodal / Human and embodied modeling
Harmonizing the object recognition strategies of deep neural networks with humans
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 neural harmonizer loss (align saliency maps with human ClickMe maps) to training | ViT · ClickMe human alignment | alignment with humans: 28.7 -> 72.6 | Helped |
| add neural harmonizer loss to training | ViT · ImageNet | top-1 accuracy: +4% | Helped |
| add neural harmonizer loss to training | EfficientNetB0 · ImageNet | top-1 accuracy | No clear effect |
| move to higher-accuracy DNNs beyond DenseNet121 (ConvNext 85.8% top-1) | 84 ImageNet DNNs · ClickMe human alignment | alignment with humans: 38 -> 22 | Hurt |
Closest papers by meaning
- Performance-optimized deep neural networks are evolving into worse models of inferotemporal visual cortex (NeurIPS 2023 · similarity 0.92)
- Evaluating alignment between humans and neural network representations in image-based learning tasks (NeurIPS 2024 · similarity 0.90)
- The 3D-PC: a benchmark for visual perspective taking in humans and machines (ICLR 2025 · similarity 0.89)
- L-WISE: Boosting Human Visual Category Learning Through Model-Based Image Selection and Enhancement (ICLR 2025 · similarity 0.89)
- Contour Integration Underlies Human-Like Vision (ICML 2025 · similarity 0.88)
- Intriguing Properties of Generative Classifiers (ICLR 2024 · 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.