# Encoder Winners Do Not Reliably Transfer Across VLA Backbone Scale: A Frozen-Backbone Grafting Diagnostic

arXiv: https://arxiv.org/abs/2606.14153 · HTML: https://mlsanity.com/paper/2606.14153/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 14 of its 17 references that are in the map.
- First posted on arXiv 12 Jun 2026.

## Builds on (most cited first, 14 of 14)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Layer Normalization](https://mlsanity.com/paper/1607.06450/) (2016)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [OpenVLA: An Open-Source Vision-Language-Action Model](https://mlsanity.com/paper/2406.09246/) (2024)
- [π0.5: a Vision-Language-Action Model with Open-World Generalization](https://mlsanity.com/paper/2504.16054/) (2025)
- [LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning](https://mlsanity.com/paper/2306.03310/) (2023)
- [Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success](https://mlsanity.com/paper/2502.19645/) (2025)
- [PaliGemma: A versatile 3B VLM for transfer](https://mlsanity.com/paper/2407.07726/) (2024)
- [LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models](https://mlsanity.com/paper/2510.13626/) (2025)
- [SmolVLM: Redefining small and efficient multimodal models](https://mlsanity.com/paper/2504.05299/) (2025)
- [FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization](https://mlsanity.com/paper/2303.14189/) (2023)
- [RepViT: Revisiting Mobile CNN From ViT Perspective](https://mlsanity.com/paper/2307.09283/) (2023)
- [Theia: Distilling Diverse Vision Foundation Models for Robot Learning](https://mlsanity.com/paper/2407.20179/) (2024)
- [VLA-0: Building State-of-the-Art VLAs with Zero Modification](https://mlsanity.com/paper/2510.13054/) (2025)

## Abstract (arXiv)

Vision-language-action (VLA) policies typically inherit their vision encoder from upstream VLM releases, but it is unclear whether an encoder choice validated on a small VLA transfers to a larger backbone. We introduce a frozen-backbone grafting diagnostic: the vision tower of a released VLA is replaced by a candidate encoder under a fixed protocol (adaptive average pooling, LayerNorm, and a single trainable linear projector), with the language model and action expert frozen. Across four encoders, two LIBERO suites, two backbones (SmolVLA-450M and π0.5-3.3B), and two-to-three seeds per cell (40 main grafting runs plus native, LoRA, pooling, and zero-/shuffled-image controls, all scored by offline action MSE), the small-backbone winner does not reliably select the large-backbone top tier: SigLIP is best on SmolVLA across both suites, while on π0.5 DINOv2-small leads the spatial suite and the object suite is a seed-sensitive near-tie band; three of the four backbone-suite comparisons (and 11 of 12 seed-level cells) support backbone-dependent rankings. The grafting wrapper is itself non-neutral with opposite sign across backbones (+45-56% MSE on the SmolVLA native tower, -50-52% on π0.5), so all conclusions are conditional on the fixed grafting protocol. We position frozen grafting as a cheap target-backbone diagnostic to run before committing to an encoder at scale, not as a closed-loop deployment claim.