# Low-Pass Filtering Improves Behavioral Alignment of Vision Models

arXiv: https://arxiv.org/abs/2602.13859 · HTML: https://mlsanity.com/paper/2602.13859/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 30 of its 59 references that are in the map.
- First posted on arXiv 14 Feb 2026.

## Builds on (most cited first, 30 of 30)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding](https://mlsanity.com/paper/2205.11487/) (2022)
- [LAION-5B: An open large-scale dataset for training next generation image-text models](https://mlsanity.com/paper/2210.08402/) (2022)
- [ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness](https://mlsanity.com/paper/1811.12231/) (2018)
- [On the Spectral Bias of Neural Networks](https://mlsanity.com/paper/1806.08734/) (2018)
- [Self-training with Noisy Student improves ImageNet classification](https://mlsanity.com/paper/1911.04252/) (2019)
- [Reproducible scaling laws for contrastive language-image learning](https://mlsanity.com/paper/2212.07143/) (2022)
- [Big Transfer (BiT): General Visual Representation Learning](https://mlsanity.com/paper/1912.11370/) (2019)
- [Scaling Vision Transformers to 22 Billion Parameters](https://mlsanity.com/paper/2302.05442/) (2023)
- [A Neural Algorithm of Artistic Style](https://mlsanity.com/paper/1508.06576/) (2015)
- [A Fourier Perspective on Model Robustness in Computer Vision](https://mlsanity.com/paper/1906.08988/) (2019)
- [Billion-scale semi-supervised learning for image classification](https://mlsanity.com/paper/1905.00546/) (2019)
- [Partial success in closing the gap between human and machine vision](https://mlsanity.com/paper/2106.07411/) (2021)
- [Getting aligned on representational alignment](https://mlsanity.com/paper/2310.13018/) (2023)
- [Beyond accuracy: quantifying trial-by-trial behaviour of CNNs and humans by measuring error consistency](https://mlsanity.com/paper/2006.16736/) (2020)
- [Shape-Texture Debiased Neural Network Training](https://mlsanity.com/paper/2010.05981/) (2020)
- [Permuted AdaIN: Reducing the Bias Towards Global Statistics in Image Classification](https://mlsanity.com/paper/2010.05785/) (2020)
- [Intriguing properties of generative classifiers](https://mlsanity.com/paper/2309.16779/) (2023)
- [The neuroconnectionist research programme](https://mlsanity.com/paper/2209.03718/) (2022)
- [Neither hype nor gloom do DNNs justice](https://mlsanity.com/paper/2312.05355/) (2023)
- [Improving neural network representations using human similarity judgments](https://mlsanity.com/paper/2306.04507/) (2023)
- [Spatial-frequency channels, shape bias, and adversarial robustness](https://mlsanity.com/paper/2309.13190/) (2023)
- [Contrast Sensitivity Functions in Autoencoders](https://mlsanity.com/paper/2103.00481/) (2021)
- [Adopting a human developmental visual diet yields robust, shape-based AI vision](https://mlsanity.com/paper/2507.03168/) (2025)
- [Increasing Shape Bias in ImageNet-Trained Networks Using Transfer Learning and Domain-Adversarial Methods](https://mlsanity.com/paper/1907.12892/) (2019)
- [LAION-C: An Out-of-Distribution Benchmark for Web-Scale Vision Models](https://mlsanity.com/paper/2506.16950/) (2025)
- [Quantifying Uncertainty in Error Consistency: Towards Reliable Behavioral Comparison of Classifiers](https://mlsanity.com/paper/2507.06645/) (2025)

## Built on by (most cited first, 1 of 1)

- [Not Too Generative, Not Too Discriminative: The Human Alignment Sweet Spot](https://mlsanity.com/paper/2605.23819/) (2026)

## Abstract (arXiv)

Despite their impressive performance on computer vision benchmarks, Deep Neural Networks (DNNs) still fall short of adequately modeling human visual behavior, as measured by error consistency and shape bias. Recent work hypothesized that behavioral alignment can be drastically improved through generative -- rather than discriminative -- classifiers, with far-reaching implications for models of human vision. Here, we instead show that the increased alignment of generative models can be largely explained by a seemingly innocuous resizing operation in the generative model which effectively acts as a low-pass filter. In a series of controlled experiments, we show that removing high-frequency spatial information from discriminative models like CLIP drastically increases their behavioral alignment. Simply blurring images at test-time -- rather than training on blurred images -- achieves a new state-of-the-art score on the model-vs-human benchmark, halving the current alignment gap between DNNs and human observers. Furthermore, low-pass filters are likely optimal, which we demonstrate by directly optimizing filters for alignment. To contextualize the performance of optimal filters, we compute the frontier of all possible pareto-optimal solutions to the benchmark, which was formerly unknown. We explain our findings by observing that the frequency spectrum of optimal Gaussian filters roughly matches the spectrum of band-pass filters implemented by the human visual system. We show that the contrast sensitivity function, describing the inverse of the contrast threshold required for humans to detect a sinusoidal grating as a function of spatiotemporal frequency, is approximated well by Gaussian filters of the specific width that also maximizes error consistency.