# EgoGazeLite: On-Device Egocentric Gaze Prediction for Token-Efficient Multimodal LLM Video Input

arXiv: https://arxiv.org/abs/2608.15614 · HTML: https://mlsanity.com/paper/2608.15614/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 16 of its 28 references that are in the map.
- First posted on arXiv 16 Aug 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 16 of 16)

- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [Ego-Exo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectives](https://mlsanity.com/paper/2311.18259/) (2023)
- [Project Aria: A New Tool for Egocentric Multi-Modal AI Research](https://mlsanity.com/paper/2308.13561/) (2023)
- [Predicting Gaze in Egocentric Video by Learning Task-dependent Attention Transition](https://mlsanity.com/paper/1803.09125/) (2018)
- [In the Eye of Transformer: Global-Local Correlation for Egocentric Gaze Estimation](https://mlsanity.com/paper/2208.04464/) (2022)
- [In the Eye of MLLM: Benchmarking Egocentric Video Intent Understanding with Gaze-Guided Prompting](https://mlsanity.com/paper/2509.07447/) (2025)
- [Multimodal Integration of Human-Like Attention in Visual Question Answering](https://mlsanity.com/paper/2109.13139/) (2021)
- [VQA-MHUG: A Gaze Dataset to Study Multimodal Neural Attention in Visual Question Answering](https://mlsanity.com/paper/2109.13116/) (2021)
- [GazeGPT: Augmenting Human Capabilities using Gaze-contingent Contextual AI for Smart Eyewear](https://mlsanity.com/paper/2401.17217/) (2024)
- [GazeLLM: Multimodal LLMs incorporating Human Visual Attention](https://mlsanity.com/paper/2504.00221/) (2025)
- [G-VOILA: Gaze-Facilitated Information Querying in Daily Scenarios](https://mlsanity.com/paper/2405.07652/) (2024)
- [Eye Gaze Tells You Where to Compute: Gaze-Driven Efficient VLMs](https://mlsanity.com/paper/2509.16476/) (2025)

## Abstract (arXiv)

The use of multimodal LLMs (MLLMs) for egocentric video understanding with wearable devices is constrained by the token budget. Memory and compute cost scale with the number of visual tokens, and high-resolution video quickly becomes expensive to transmit and process at scale. Prior work (GazeLLM) addresses this by cropping the video around the camera wearer's gaze. This reduces the number of visual tokens by about tenfold while maintaining or improving the quality of full-resolution descriptions. However, this compression strategy depends on dedicated eye-tracking hardware, which is unavailable on consumer smart glasses. Building a software-only substitute poses a joint constraint: the predictor must be accurate enough to preserve downstream description quality, yet light enough to run on-device, within the power and compute budget of a smartphone. We address this with EgoGazeLite, a lightweight dual-process gaze predictor for egocentric video. Across two MLLMs, three automated metrics, and two LLM judges, predicted-gaze crops show no significant difference from ground-truth-gaze crops. Equivalence is confirmed in all ten cases. EgoGazeLite achieves this at 15.7M parameters, 6.71 GFLOPs, and runs the full gaze-and-crop pipeline end-to-end in real time (21.6 ms/frame) on consumer accelerator hardware. Together, these results remove the need for eye-tracking hardware for token-efficient, gaze-conditioned egocentric video understanding with MLLMs.