# Retrieval-Augmented Gaussian Avatars: Improving Expression Generalization

arXiv: https://arxiv.org/abs/2603.08645 · HTML: https://mlsanity.com/paper/2603.08645/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Multimodal RAG, one of 15 directions in Retrieval-augmented generation.
- Builds on 20 of its 29 references that are in the map.
- First posted on arXiv 9 Mar 2026.

## Placed in

- Retrieval-augmented generation / [Multimodal RAG](https://mlsanity.com/rag/multimodal-rag/)

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

- [3D Gaussian Splatting for Real-Time Radiance Field Rendering](https://mlsanity.com/paper/2308.04079/) (2023)
- [Face2Face: Real-time Face Capture and Reenactment of RGB Videos](https://mlsanity.com/paper/2007.14808/) (2020)
- [Learning an Animatable Detailed 3D Face Model from In-The-Wild Images](https://mlsanity.com/paper/2012.04012/) (2020)
- [Dynamic Neural Radiance Fields for Monocular 4D Facial Avatar Reconstruction](https://mlsanity.com/paper/2012.03065/) (2020)
- [Deep Video Portraits](https://mlsanity.com/paper/1805.11714/) (2018)
- [GaussianAvatars: Photorealistic Head Avatars with Rigged 3D Gaussians](https://mlsanity.com/paper/2312.02069/) (2023)
- [Neural Head Avatars from Monocular RGB Videos](https://mlsanity.com/paper/2112.01554/) (2021)
- [NeRSemble: Multi-view Radiance Field Reconstruction of Human Heads](https://mlsanity.com/paper/2305.03027/) (2023)
- [PointAvatar: Deformable Point-based Head Avatars from Videos](https://mlsanity.com/paper/2212.08377/) (2022)
- [Instant Volumetric Head Avatars](https://mlsanity.com/paper/2211.12499/) (2022)
- [Real3D-Portrait: One-shot Realistic 3D Talking Portrait Synthesis](https://mlsanity.com/paper/2401.08503/) (2024)
- [Generalizable and Animatable Gaussian Head Avatar](https://mlsanity.com/paper/2410.07971/) (2024)
- [MonoGaussianAvatar: Monocular Gaussian Point-based Head Avatar](https://mlsanity.com/paper/2312.04558/) (2023)
- [AvatarMAV: Fast 3D Head Avatar Reconstruction Using Motion-Aware Neural Voxels](https://mlsanity.com/paper/2211.13206/) (2022)
- [Avat3r: Large Animatable Gaussian Reconstruction Model for High-fidelity 3D Head Avatars](https://mlsanity.com/paper/2502.20220/) (2025)
- [CAP4D: Creating Animatable 4D Portrait Avatars with Morphable Multi-View Diffusion Models](https://mlsanity.com/paper/2412.12093/) (2024)
- [LAM: Large Avatar Model for One-shot Animatable Gaussian Head](https://mlsanity.com/paper/2502.17796/) (2025)
- [Demystifying Inter-Class Disentanglement](https://mlsanity.com/paper/1906.11796/) (2019)
- [Gaussian Eigen Models for Human Heads](https://mlsanity.com/paper/2407.04545/) (2024)
- [SEGA: Drivable 3D Gaussian Head Avatar from a Single Image](https://mlsanity.com/paper/2504.14373/) (2025)

## Abstract (arXiv)

Template-free animatable head avatars can achieve high visual fidelity by learning expression-dependent facial deformation directly from a subject's capture, avoiding parametric face templates and hand-designed blendshape spaces. However, since learned deformation is supervised only by the expressions observed for a single identity, these models suffer from limited expression coverage and often struggle when driven by motions that deviate from the training distribution. We introduce RAF (Retrieval-Augmented Faces), a simple training-time augmentation designed for template-free head avatars that learn deformation from data. RAF constructs a large unlabeled expression bank and, during training, replaces a subset of the subject's expression features with nearest-neighbor expressions retrieved from this bank while still reconstructing the subject's original frames. This exposes the deformation field to a broader range of expression conditions, encouraging stronger identity-expression decoupling and improving robustness to expression distribution shift without requiring paired cross-identity data, additional annotations, or architectural changes. We further analyze how retrieval augmentation increases expression diversity and validate retrieval quality with a user study showing that retrieved neighbors are perceptually closer in expression and pose. Experiments on the NeRSemble benchmark demonstrate that RAF consistently improves expression fidelity over the baseline, in both self-driving and cross-driving scenarios.