# EVA: Bridging Performance and Human Alignment in Hard-Attention Vision Models for Image Classification

arXiv: https://arxiv.org/abs/2603.27340 · HTML: https://mlsanity.com/paper/2603.27340/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 39 references that are in the map.
- First posted on arXiv 28 Mar 2026.

## Builds on (most cited first, 20 of 20)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [Show, Attend and Tell: Neural Image Caption Generation with Visual Attention](https://mlsanity.com/paper/1502.03044/) (2015)
- [Recurrent Models of Visual Attention](https://mlsanity.com/paper/1406.6247/) (2014)
- [TinyViT: Fast Pretraining Distillation for Small Vision Transformers](https://mlsanity.com/paper/2207.10666/) (2022)
- [Convolutional Neural Networks as a Model of the Visual System: Past, Present, and Future](https://mlsanity.com/paper/2001.07092/) (2020)
- [DeepGaze IIE: Calibrated prediction in and out-of-domain for state-of-the-art saliency modeling](https://mlsanity.com/paper/2105.12441/) (2021)
- [Predicting Goal-directed Human Attention Using Inverse Reinforcement Learning](https://mlsanity.com/paper/2005.14310/) (2020)
- [Gazeformer: Scalable, Effective and Fast Prediction of Goal-Directed Human Attention](https://mlsanity.com/paper/2303.15274/) (2023)
- [Exploring Human-like Attention Supervision in Visual Question Answering](https://mlsanity.com/paper/1709.06308/) (2017)
- [Saccader: Improving Accuracy of Hard Attention Models for Vision](https://mlsanity.com/paper/1908.07644/) (2019)
- [Unifying Top-down and Bottom-up Scanpath Prediction Using Transformers](https://mlsanity.com/paper/2303.09383/) (2023)
- [Target-absent Human Attention](https://mlsanity.com/paper/2207.01166/) (2022)
- [GazeXplain: Learning to Predict Natural Language Explanations of Visual Scanpaths](https://mlsanity.com/paper/2408.02788/) (2024)
- [Trends, Applications, and Challenges in Human Attention Modelling](https://mlsanity.com/paper/2402.18673/) (2024)
- [Modeling Human Gaze Behavior with Diffusion Models for Unified Scanpath Prediction](https://mlsanity.com/paper/2507.23021/) (2025)
- [Active Sensing with Predictive Coding and Uncertainty Minimization](https://mlsanity.com/paper/2307.00668/) (2023)
- [Gaze-Guided Learning: Avoiding Shortcut Bias in Visual Classification](https://mlsanity.com/paper/2504.05583/) (2025)
- [Emergence of Fixational and Saccadic Movements in a Multi-Level Recurrent Attention Model for Vision](https://mlsanity.com/paper/2505.13191/) (2025)
- [Debiasing Central Fixation Confounds Reveals a Peripheral "Sweet Spot" for Human-like Scanpaths in Hard-Attention Vision](https://mlsanity.com/paper/2602.14834/) (2026)

## Abstract (arXiv)

Optimizing vision models purely for classification accuracy can impose an alignment tax, degrading human-like scanpaths and limiting interpretability. We introduce EVA, a neuroscience-inspired hard-attention mechanistic testbed that makes the performance-human-likeness trade-off explicit and adjustable. EVA samples a small number of sequential glimpses using a minimal fovea-periphery representation with CNN-based feature extractor and integrates variance control and adaptive gating to stabilize and regulate attention dynamics. EVA is trained with the standard classification objective without gaze supervision. On CIFAR-10 with dense human gaze annotations, EVA improves scanpath alignment under established metrics such as DTW, NSS, while maintaining competitive accuracy. Ablations show that CNN-based feature extraction drives accuracy but suppresses human-likeness, whereas variance control and gating restore human-aligned trajectories with minimal performance loss. We further validate EVA's scalability on ImageNet-100 and evaluate scanpath alignment on COCO-Search18 without COCO-Search18 gaze supervision or finetuning, where EVA yields human-like scanpaths on natural scenes without additional training. Overall, EVA provides a principled framework for trustworthy, human-interpretable active vision.