# Test-Time Adaptation with Online Personalized Energy-Based Cache for Fine-Grained Video Expression Recognition

arXiv: https://arxiv.org/abs/2608.06467 · HTML: https://mlsanity.com/paper/2608.06467/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 37 references that are in the map.
- First posted on arXiv 6 Aug 2026.

## Builds on (most cited first, 19 of 19)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [CLIP-Adapter: Better Vision-Language Models with Feature Adapters](https://mlsanity.com/paper/2110.04544/) (2021)
- [Your Classifier is Secretly an Energy Based Model and You Should Treat it Like One](https://mlsanity.com/paper/1912.03263/) (2019)
- [Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm](https://mlsanity.com/paper/2110.05208/) (2021)
- [Expanding Language-Image Pretrained Models for General Video Recognition](https://mlsanity.com/paper/2208.02816/) (2022)
- [Robust Test-Time Adaptation in Dynamic Scenarios](https://mlsanity.com/paper/2303.13899/) (2023)
- [Diverse Data Augmentation with Diffusions for Effective Test-time Prompt Tuning](https://mlsanity.com/paper/2308.06038/) (2023)
- [Source-free Domain Adaptation via Avatar Prototype Generation and Adaptation](https://mlsanity.com/paper/2106.15326/) (2021)
- [CLIPood: Generalizing CLIP to Out-of-Distributions](https://mlsanity.com/paper/2302.00864/) (2023)
- [C-TPT: Calibrated Test-Time Prompt Tuning for Vision-Language Models via Text Feature Dispersion](https://mlsanity.com/paper/2403.14119/) (2024)
- [VDM-DA: Virtual Domain Modeling for Source Data-free Domain Adaptation](https://mlsanity.com/paper/2103.14357/) (2021)
- [A Closer Look at the Robustness of Contrastive Language-Image Pre-Training (CLIP)](https://mlsanity.com/paper/2402.07410/) (2024)
- [WATT: Weight Average Test-Time Adaptation of CLIP](https://mlsanity.com/paper/2406.13875/) (2024)
- [Enhancing Zero-Shot Facial Expression Recognition by LLM Knowledge Transfer](https://mlsanity.com/paper/2405.19100/) (2024)
- [Disentangled Source-Free Personalization for Facial Expression Recognition with Neutral Target Data](https://mlsanity.com/paper/2503.20771/) (2025)
- [Advancing Reliable Test-Time Adaptation of Vision-Language Models under Visual Variations](https://mlsanity.com/paper/2507.09500/) (2025)
- [Test-Time Adaptation via Cache Personalization for Facial Expression Recognition in Videos](https://mlsanity.com/paper/2603.21309/) (2026)
- [MuSACo: Multimodal Subject-Specific Selection and Adaptation for Expression Recognition with Co-Training](https://mlsanity.com/paper/2508.12522/) (2025)
- [CLIP-AUTT: Test-Time Personalization with Action Unit Prompting for Fine-Grained Video Emotion Recognition](https://mlsanity.com/paper/2603.27999/) (2026)

## Abstract (arXiv)

Facial expression recognition (FER) in videos is challenging because models must identify subtle, temporally evolving affective states that vary across individuals. Although vision-language models provide transferable visual-semantic representations, models trained on subject-independent data often degrade under subject-specific distribution shifts at inference time. Existing test-time adaptation (TTA) methods commonly update model parameters during inference, increasing computational cost and latency. Cache-based methods avoid parameter updates, but they usually require enough target samples to form reliable class prototypes, which is difficult early in adaptation and for rarely observed classes. We introduce Energy-Based Cache Personalization (EB-CaP), a subject-based online TTA method for video FER that generates class-specific prototypes personalized to each target video. EB-CaP uses a lightweight energy-based model to sample prototypes from the current unlabeled video and populate a personalized cache online, without accumulating large amounts of target data or storing diverse source prototypes. Its energy function relies only on pretrained CLIP: similarities between the target video embedding and class text embeddings guide prototype sampling. In parallel, positive and negative caches store reliable and uncertain target embeddings. An adaptive entropy gate controls cache updates according to the evolving confidence distribution, while a diversity gate limits redundant samples. Final predictions combine cache-derived scores with the current CLIP scores. Experiments on BioVid, StressID, and BAH show that EB-CaP outperforms state-of-the-art TTA methods while maintaining low computational and memory overhead. Code is available at https://github.com/MasoumehSharafi/EB-CaP.