Abstract
DiG-Flow enhances VLA models' robustness by using geometric regularization to align observation and action embeddings, improving performance on complex tasks and with limited data.
Vision-Language-Action (VLA) models trained with flow matching have demonstrated impressive capabilities on robotic manipulation tasks. However, their performance often degrades under distribution shift and on complex multi-step tasks, suggesting that the learned representations may not robustly capture task-relevant semantics. We introduce DiG-Flow, a principled framework that enhances VLA robustness through geometric regularization. Our key insight is that the distributional discrepancy between observation and action embeddings provides a meaningful geometric signal: lower transport cost indicates compatible representations, while higher cost suggests potential misalignment. DiG-Flow computes a discrepancy measure between empirical distributions of observation and action embeddings, maps it to a modulation weight via a monotone function, and applies residual updates to the observation embeddings before flow matching. Crucially, this intervention operates at the representation level without modifying the flow matching path or target vector field. We provide theoretical guarantees showing that discrepancy-guided training provably decreases the training objective, and that guided inference refinement converges with contraction. Empirically, DiG-Flow integrates into existing VLA architectures with negligible overhead and consistently improves performance, with particularly pronounced gains on complex multi-step tasks and under limited training data.
Community
DiG-Flow is a plug-and-play module for flow-matching based VLAs that rebalances control between the autoregressive foundation model and the flow expert. It embeds model inputs and flow outputs into a unified discrepancy space and uses this signal to gate the flow path, preventing shortcut transports that bypass the pretrained model and steering the expert toward more general, robust actions. DiG-Flow integrates seamlessly into diverse VLA architectures.
Project Page: https://beingbeyond.github.io/DiG-Flow
arXiv: https://arxiv.org/abs/2512.01715
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- QDepth-VLA: Quantized Depth Prediction as Auxiliary Supervision for Vision-Language-Action Models (2025)
- MAPS: Preserving Vision-Language Representations via Module-Wise Proximity Scheduling for Better Vision-Language-Action Generalization (2025)
- ActDistill: General Action-Guided Self-Derived Distillation for Efficient Vision-Language-Action Models (2025)
- Evo-1: Lightweight Vision-Language-Action Model with Preserved Semantic Alignment (2025)
- VLA Models Are More Generalizable Than You Think: Revisiting Physical and Spatial Modeling (2025)
- Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Model (2025)
- ACG: Action Coherence Guidance for Flow-based VLA models (2025)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper