# DEFLECT: Temporal Counterfactual Preference Learning for Delay-Robust Asynchronous VLAs

arXiv: https://arxiv.org/abs/2605.19294 · HTML: https://mlsanity.com/paper/2605.19294/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Efficient VLA inference and adaptation, one of 13 directions in Vision-language-action models.
- Builds on 22 of its 47 references that are in the map.
- First posted on arXiv 19 May 2026.

## Placed in

- Vision-language-action models / [Efficient VLA inference and adaptation](https://mlsanity.com/vla/efficient-vla-inference-and-adaptation/)

## Builds on (most cited first, 22 of 22)

- [Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware](https://mlsanity.com/paper/2304.13705/) (2023)
- [Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success](https://mlsanity.com/paper/2502.19645/) (2025)
- [FAST: Efficient Action Tokenization for Vision-Language-Action Models](https://mlsanity.com/paper/2501.09747/) (2025)
- [Flow-GRPO: Training Flow Matching Models via Online RL](https://mlsanity.com/paper/2505.05470/) (2025)
- [WorldVLA: Towards Autoregressive Action World Model](https://mlsanity.com/paper/2506.21539/) (2025)
- [TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulation](https://mlsanity.com/paper/2409.12514/) (2024)
- [X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Model](https://mlsanity.com/paper/2510.10274/) (2025)
- [Real-Time Execution of Action Chunking Flow Policies](https://mlsanity.com/paper/2506.07339/) (2025)
- [HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Model](https://mlsanity.com/paper/2503.10631/) (2025)
- [Steering Your Diffusion Policy with Latent Space Reinforcement Learning](https://mlsanity.com/paper/2506.15799/) (2025)
- [DeeR-VLA: Dynamic Inference of Multimodal Large Language Models for Efficient Robot Execution](https://mlsanity.com/paper/2411.02359/) (2024)
- [Flow Matching Policy Gradients](https://mlsanity.com/paper/2507.21053/) (2025)
- [ReinFlow: Fine-tuning Flow Matching Policy with Online Reinforcement Learning](https://mlsanity.com/paper/2505.22094/) (2025)
- [ManiFlow: A General Robot Manipulation Policy via Consistency Flow Training](https://mlsanity.com/paper/2509.01819/) (2025)
- [DynaGuide: Steering Diffusion Polices with Active Dynamic Guidance](https://mlsanity.com/paper/2506.13922/) (2025)
- [CO-RFT: Efficient Fine-Tuning of Vision-Language-Action Models through Chunked Offline Reinforcement Learning](https://mlsanity.com/paper/2508.02219/) (2025)
- [MP1: MeanFlow Tames Policy Learning in 1-step for Robotic Manipulation](https://mlsanity.com/paper/2507.10543/) (2025)
- [FASTER: Rethinking Real-Time Flow VLAs](https://mlsanity.com/paper/2603.19199/) (2026)
- [π_textttRL: Online RL Fine-tuning for Flow-based Vision-Language-Action Models](https://mlsanity.com/paper/2510.25889/) (2025)
- [SnapFlow: One-Step Action Generation for Flow-Matching VLAs via Progressive Self-Distillation](https://mlsanity.com/paper/2604.05656/) (2026)
- [VLP: Vision-Language Preference Learning for Embodied Manipulation](https://mlsanity.com/paper/2502.11918/) (2025)
- [Reparameterization Flow Policy Optimization](https://mlsanity.com/paper/2602.03501/) (2026)

## Built on by (most cited first, 1 of 1)

- [Reactive Real-Time Flow Policies via Asynchronous Distribution Alignment](https://mlsanity.com/paper/2609.36540/) (2026)

## Abstract (arXiv)

Vision-Language-Action (VLA) policies increasingly rely on asynchronous inference to hide large-model latency behind ongoing robot motion. While this avoids the stop-and-go behavior of synchronous action-chunk execution, it creates a prediction-execution mismatch: the next chunk is computed from a stale observation at inference start but executed only after the robot and scene have evolved. As a result, actions that fit the prediction-time state can become misaligned with the execution-time state. Existing runtime repair, behavior-cloning, and preference-alignment approaches do not directly teach the policy to resolve this stale-input mismatch. We propose DEFLECT, an offline post-training framework for delay-robust asynchronous VLAs. DEFLECT converts latency-induced mismatch into counterfactual preference supervision: a frozen reference VLA generates a preferred chunk from the future execution-time observation and a rejected chunk from the stale prediction-time observation. The trainable policy scores both chunks under the same deployment-time input, learning to favor execution-time-aligned actions while a supervised fine-tuning anchor preserves the expert action manifold. DEFLECT requires no human preference labels, reward models, online robot rollouts, architectural changes, or additional inference-time computation. Across Kinetix, LIBERO, and three real-robot tasks, DEFLECT improves delay robustness over strong asynchronous VLA baselines, raising high-latency success by up to 6.4 percentage points and achieving a 4.6 percentage-point gain at the longest delay on a real-scale VLA.