aiqbench analysis
The paper proposes a way to recover frozen vision-language-action robot policies by using a learned world model to compare the policy's own continuation against camera-reobservation and replanning options, scored by a task-progress head. The reported gain on simulated RoboTwin 2.0 tasks (13.3% to 30.0% pooled success) is substantial but comes from injected failures in simulation, with only a limited real-robot adaptation step. Readers should watch whether the approach holds up on real hardware and across more backbones, and whether the code release allows independent checks.
ForkCast: Predictive Recovery for Vision-Language-Action Policies via Counterfactual Futures
arXiv:2610.18650v1 Announce Type: new Abstract: Vision-language-action (VLA) policies perform manipulation tasks using camera images and language instructions, but object slip or visual occlusion can disrupt execution. Recovery is typically learned during policy training or delegated to a monitor that triggers a scripted repair after detecting a failure. We introduce ForkCast, a predictive recovery framework that compares counterfactual futures to determine whether and how to intervene in a frozen VLA policy. At each decision point, ForkCast evaluates three candidates: executing the policy's nominal action chunk (Commit), adjusting the camera viewpoint (Reobserve), and replanning the arm trajectory (Replan). A candidate-conditioned world model forecasts the latent outcome of each candidate at a common future horizon, and a shared task-progress head scores the predicted outcomes. By evaluating the policy's own continuation alongside the repair candidates, a single comparison based on task progress determines both whether and how to intervene, without a failure detector, an intervention threshold, or updates to the VLA backbone. On five simulated RoboTwin 2.0 tasks with injected occlusion and object slip, ForkCast increases pooled success from 13.3% for the frozen $\pi_{0.5}$ policy to 30.0% and transfers to two additional VLA backbones without updating ForkCast. After adaptation on real-robot candidate rollouts, ForkCast also improves success over the base policy under both clean and visibility-limited conditions. Code will be released to benefit the research community.
Abstract:Vision-language-action (VLA) policies perform manipulation tasks using camera images and language instructions, but object slip or visual occlusion can disrupt execution. Recovery is typically learned during policy training or delegated to a monitor that triggers a scripted repair after detecting a failure. We introduce ForkCast, a predictive recovery framework that compares counterfactual futures to determine whether and how to intervene in a frozen VLA policy. At each decision point, ForkCast evaluates three candidates: executing the policy's nominal action chunk (Commit), adjusting the camera viewpoint (Reobserve), and replanning the arm trajectory (Replan). A candidate-conditioned world model forecasts the latent outcome of each candidate at a common future horizon, and a shared task-progress head scores the predicted outcomes. By evaluating the policy's own continuation alongside the repair candidates, a single comparison based on task progress determines both whether and how to intervene, without a failure detector, an intervention threshold, or updates to the VLA backbone. On five simulated RoboTwin 2.0 tasks with injected occlusion and object slip, ForkCast increases pooled success from 13.3% for the frozen $\pi_{0.5}$ policy to 30.0% and transfers to two additional VLA backbones without updating ForkCast. After adaptation on real-robot candidate rollouts, ForkCast also improves success over the base policy under both clean and visibility-limited conditions. Code will be released to benefit the research community.
Robotics (cs.RO)
arXiv:2610.18650 [cs.RO]
(or arXiv:2610.18650v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.18650 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 310 words
Related coverage
Building a Foundation Stack for General-Purpose Robots
IEEE Spectrum · Jul 13, 2026
Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
AWS · Sep 4, 2026
Runway’s WorldPrompt and the Engineering of Real-Time Worlds
Latent Space · Sep 25, 2026
IMPACT-VLA: Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies
arXiv · Sep 16, 2026