MAPS: Preserving Vision-Language Representations via Module-Wise Proximity Scheduling for Better Vision-Language-Action Generalization

*Equal contribution
Georgia Institute of Technology
CVPR 2026

The MAPS Framework

Motivation: Why does naive fine-tuning fail?

VLA models initialize from powerful VLMs pretrained on web-scale data, leveraging their rich visual and semantic priors before fine-tuning on robotics datasets to learn action policies. However, fine-tuning on sparse, task-specific robot data induces spurious correlations and overfitting, causing catastrophic forgetting of spatial reasoning, world knowledge, and linguistic grounding. The core challenge is balancing the tradeoff between action adaptation and preservation of pretrained generalization.

Freezing visual encoders preserves pretrained perception but limits adaptation. Uniform regularization constrains all layers equally, ignoring the fact that different components encode distinct priors and vary in their sensitivity to fine-tuning. Some modules must stay near their pretrained state to preserve structural knowledge; others require greater flexibility to adapt task-specific grounding.

SimplerEnv preliminary results
LIBERO preliminary results and L2 distance from pretrained weights

A Module Hierarchy for VLAs

Through systematic analysis of freezing configurations across DINOv2, SigLIP, early language layers, and late language layers, we identify an empirical hierarchy of how much each component should be preserved:

DINOv2 > SigLIP > Early Language > Late Language

DINOv2 contributes essential depth perception and geometric priors for manipulation. SigLIP provides strong vision-language alignment. Late language layers drive task performance and benefit from more aggressive adaptation to the action space. Freezing any single component in isolation introduces task-dependent inductive biases that degrade performance across benchmarks — freezing alone is not a reliable path to generalization.

Module-Wise Proximity Scheduling (MAPS)

MAPS Framework

MAPS assigns a linearly decaying proximity weight $\lambda_k = \lambda_{\max} \cdot \left(1 - \frac{k-1}{|L|-1}\right)$ to each module — strongest for early visual layers (DINOv2, SigLIP) and weakest for the deepest language layers — anchoring pretrained perception while action-oriented language layers adapt more freely. Modules with no pretrained parameters (action head, proprioceptive projector) are assigned $\lambda = 0$ and undergo full fine-tuning.

Real-Robot Demonstrations (Franka)

MiniVLA-OFT baseline vs. +MAPS across four tasks, in both ID and OOD settings.

Task 1: Can

In-Distribution

Coke can

Baseline

+ MAPS (Ours)

Out-of-Distribution

Red Bull can (novel object, color & geometry)

Baseline

+ MAPS (Ours)

Task 2: Block

In-Distribution

Stack blue block on green block

Baseline

+ MAPS (Ours)

Out-of-Distribution

Blue block elevated on a third block

Baseline

+ MAPS (Ours)

Task 3: Cup

In-Distribution

Stack orange cup onto green cup

Baseline

+ MAPS (Ours)

Out-of-Distribution

Yellow cup onto blue cup (novel colors)

Baseline

+ MAPS (Ours)

Task 4: Laptop

In-Distribution

Close Alienware laptop

Baseline

+ MAPS (Ours)

Out-of-Distribution

MacBook (novel class, shape & joint stiffness)

Baseline

+ MAPS (Ours)

Results

Real Robot (Franka)

Franka real-robot results

Franka Emika Panda. MAPS boosts ID success 40%→72.5% and OOD 22.5%→52.5% across four tasks (can, block, cup, laptop).

SimplerEnv

SimplerEnv results

SimplerEnv (BridgeData V2). MAPS matches ID performance while improving OOD generalization by up to 20% on MiniVLA-OFT, surpassing RT-1-X, Octo, and π₀ on OOD tasks.

CALVIN ABC→D and LIBERO

CALVIN results
LIBERO results

CALVIN ABC→D. MAPS improves average sequence length by +0.7 (MiniVLA-OFT) / +0.1 (OpenVLA-OFT).   LIBERO. With no external VLA pretraining, MAPS delivers consistent OOD gains of 1–10% across all four backbones on unseen task splits.

BibTeX

@inproceedings{huang2026maps,
  title     = {{MAPS}: Preserving Vision-Language Representations via Module-Wise Proximity Scheduling for Better {VLA} Generalization},
  author    = {Huang, Chengyue and Zhang, Mellon M. and Azarcon, Robert and Chou, Glen and Kira, Zsolt},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year      = {2026},
}