# Efficient Universal Perception Encoder

Source: [Efficient Universal Perception Encoder](<https://www.sota2.com/research/paper/69c36be80636e5bf4afc6338>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-02T05:00:12.805Z

Authors: Chenchen Zhu, Saksham Suri, Cijo Jose, Maxime Oquab, Marc Szafraniec, Wei Wen, Yunyang Xiong, Patrick Labatut, Piotr Bojanowski, Raghuraman Krishnamoorthi, Vikas Chandra

Published: 2026-03-23T17:50:19.000Z

## Abstract

Running AI models on smart edge devices can unlock versatile user experiences, but presents challenges due to limited compute and the need to handle multiple tasks simultaneously. This requires a vision encoder with small size but powerful and versatile representations. We present our method, Efficient Universal Perception Encoder (EUPE), which offers both inference efficiency and universally good representations for diverse downstream tasks. We achieve this by distilling from multiple domain-expert foundation vision encoders. Unlike previous agglomerative methods that directly scale down from multiple teachers to an efficient encoder, we demonstrate the importance of first scaling up to a large proxy teacher and then scaling down from this single teacher. Experiments show that EUPE achieves on-par or better performance than individual domain experts of the same size on diverse task domains and also outperforms previous agglomerative encoders. We release the full family of EUPE models and the code to foster future research.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [KNN classification on ImageNet-1k (val)](<https://www.sota2.com/research/sota/knn-classification-on-imagenet-1k-val.md>): KNN Classification; ImageNet-1k (val); Top-1 Accuracy: 84.1.

- [Zero-shot Image Classification on ImageNet-1K](<https://www.sota2.com/research/sota/zero-shot-image-classification-on-imagenet-1k.md>): Zero-shot Image Classification; ImageNet-1K; Top-1 Accuracy: 79.7.

- [Vision-Language Perception on MME Perception](<https://www.sota2.com/research/sota/vision-language-perception-on-mme-perception.md>): Vision-Language Perception; MME Perception; Perception Score: 1374.5.

- [Open-Vocabulary Semantic Segmentation on ORAD-3D](<https://www.sota2.com/research/sota/open-vocabulary-semantic-segmentation-on-orad-3d.md>): Open Vocabulary Semantic Segmentation; ORAD-3D; mIoU: 19.85.

- [Visual Question Answering on POPE](<https://www.sota2.com/research/sota/visual-question-answering-on-pope.md>): Visual Question Answering; POPE; Accuracy: 85.9.

- [Visual Question Answering on GQA](<https://www.sota2.com/research/sota/visual-question-answering-on-gqa.md>): Visual Question Answering; GQA; Accuracy: 67.3.

- [Semantic Segmentation on ADE20K (mIoU)](<https://www.sota2.com/research/sota/semantic-segmentation-on-ade20k-miou.md>): Semantic segmentation; ADE20K; mIoU: 52.4.

- [Depth Estimation on NYU v2](<https://www.sota2.com/research/sota/depth-estimation-on-nyu-v2.md>): Depth Estimation; NYU V2; RMSE: 0.391.

- [Visual Question Answering on RealworldQA](<https://www.sota2.com/research/sota/visual-question-answering-on-realworldqa.md>): Visual Question Answering; RealworldQA; Accuracy: 55.5.

- [Visual Question Answering on ScienceQA](<https://www.sota2.com/research/sota/visual-question-answering-on-scienceqa.md>): Visual Question Answering; ScienceQA; Accuracy: 69.7.

- [Semantic Correspondence on SPair-71k (PCK)](<https://www.sota2.com/research/sota/semantic-correspondence-on-spair-71k-pck.md>): Semantic Correspondence; SPair-71k; PCK@0.01: not recorded.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.22387>)

[PDF](<https://arxiv.org/pdf/2603.22387.pdf>)
