小模型选的视觉编码器，大模型不一定适用。
Encoder Winners Do Not Reliably Transfer Across VLA Backbone Scale: A Frozen-Backbone Grafting Diagnostic
- 固定语言和动作模块，只换视觉编码器做对比测试
- 小模型表现好的编码器在大模型上未必领先，排名会变
- 适合评估编码器跨规模迁移性，尤其在部署前做筛选
视觉-语言-动作（VLA）策略通常继承上游视觉语言模型（VLM）的视觉编码器，但尚不清楚小规模VLA验证有效的编码器是否能在更大模型上延续优势。本文提出一种冻结主干嫁接诊断方法：在固定协议（自适应平均池化、LayerNorm、单个可训练线性投影头）下，将已发布VLA的视觉塔替换为候选编码器，同时冻结语言模型和动作专家。实验涵盖四个编码器、两个LIBERO任务套件、两个主干（SmolVLA-450M 和 $π_{0.5}$-3.3B），每组2~3个随机种子（共40次主实验及原生、LoRA、池化、零图/乱序图控制）。结果显示，小主干上的优胜编码器无法可靠地在大主干上保持领先：在SmolVLA上SigLIP表现最佳，而在$π_{0.5}$上DINOv2-small在空间任务中领先，物体任务则为种子敏感的近似平局；12个种子单元中有11个支持主干依赖的排序。嫁接框架本身并非中立，对不同主干产生相反影响（在SmolVLA上增加45%-56% MSE，$π_{0.5}$上减少50%-52%），所有结论均依赖于固定嫁接协议。本文将冻结嫁接定位为低成本的预部署诊断工具，而非闭环部署依据。
原文摘要 · Abstract (English)
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 $π_{0.5}$-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 $π_{0.5}$ 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 $π_{0.5}$), 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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。