Driving · 2026-10-14

ForkCast: Predictive Recovery for Vision-Language-Action Policies via Counterfactual Futures

Chenru Wen, Huajian Zeng, Haokai Ding, Jian Kang, Anqing Duan, Dawei Zhou, Dezhen Song, Xingxing Zuo

arXiv:2610.18650PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

ForkCast is a predictive recovery method for frozen vision-language-action (VLA) policies. At each action-chunk boundary it compares counterfactual futures for three candidates—Commit (nominal continuation), Reobserve (camera viewpoint change), and Replan (planned arm correction)—using a candidate-conditioned world model and a shared task-progress head. It selects the highest-scoring candidate without a failure detector, intervention threshold, or VLA updates.

Why it matters

The authors address VLA execution failures from occlusion and object slip by selecting among nominal continuation and predefined repair primitives using predicted task progress at a common horizon. This aims to decide both whether to intervene and which repair to use, while keeping the VLA backbone frozen.

Method

  • At each decision point, ForkCast generates candidates: Commit (nominal chunk), Reobserve (probe views + re-query frozen policy), and Replan (planner-based arm correction).
  • A candidate-conditioned latent world model forecasts each candidate’s counterfactual endpoint latent at a shared horizon; a shared progress head scores predicted endpoints.
  • ForkCast executes the argmax candidate score (ties favor Commit), with Reobserve using cached action chunks and Replan consuming additional arm-control steps.

Limitation

ForkCast is bounded by the capabilities of its predefined candidates and does not add new manipulation skills to the frozen policy. In the profiled serial implementation, image-to-selection latency averages 1.439 s per decision; with candidate generation and physical viewpoint acquisition, this may limit responsiveness in rapidly changing scenes. End-to-end deployment latency and the benefits of candidate batching remain to be evaluated.

Abstract (from arXiv)

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.

Related papers

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-10-13 · Indexed by Pub-AI: 2026-10-14 UTC

preprint

ForkCast: Predictive Recovery for Vision-Language-Action Policies via Counterfactual Futures

Chenru Wen, Huajian Zeng, Haokai Ding, Jian Kang, Anqing Duan, Dawei Zhou, Dezhen Song, Xingxing Zuo

Preprint posted: 2026-10-13 · Original record ↗

arxiv
2610.18650
Explicit source versions (1)
  • v1 · 2026-10-13
  1. Preprint available

    Source date: 2026-10-13 · First detected: 2026-10-25 UTC · Historical baseline

    Source evidence (1)
Abstract · arxiv

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 $π_{0.5}$ 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.