arXiv Robotics — research abstracts· Feihong Zhang, Xi Lin, Yulong Shi, Zuxing Lu, Yanghong Mei, Zhaopeng Gu, Tingyu Yuan, Zihao Liang, Yinuo Wang, Zhaowen Li·2026-10-30 04:00 UTC· 1 days agoSignalEditorial score83

Utility-WAM: Correcting Future-Induced Action Changes

Summary

This research paper examines how predictive future contexts affect robot action execution, identifying a failure mode called 'future-induced negative transfer' that harms precision tasks. The paper introduces Utility-WAM, a framework that corrects these issues using a residual adapter, improving success rates and reducing inference latency across multiple benchmarks.

Source: arXiv Robotics — research abstracts · Read original article ↗

Loading article text…

What the source reports

Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.

Reported numbers

benchmark success points
0.8
View original evidence
hifts average benchmark success ($+0.8$ points on LIBERO, $-0.8$ on RoboTwin~2.0), while generating visual futures increases inference latency by 1.7--1.9$\times$. These aggregate metrics conceal a concentrated failure mode that we term \emph{future-induced negative transfer}: while predictive deltas benefit long-horizon tasks ($+2.4$ points on LIBERO-Long) and remain negligible on most routine manipulations, they in
Open source S3
inference latency multiplier
1.7
View original evidence
hifts average benchmark success ($+0.8$ points on LIBERO, $-0.8$ on RoboTwin~2.0), while generating visual futures increases inference latency by 1.7--1.9$\times$. These aggregate metrics conceal a concentrated failure mode that we term \emph{future-induced negative transfer}: while predictive deltas benefit long-horizon tasks ($+2.4$ points on LIBERO-Long) and remain negligible on most routine manipulations, they in
Open source S3
inference latency multiplier
1.9
View original evidence
hifts average benchmark success ($+0.8$ points on LIBERO, $-0.8$ on RoboTwin~2.0), while generating visual futures increases inference latency by 1.7--1.9$\times$. These aggregate metrics conceal a concentrated failure mode that we term \emph{future-induced negative transfer}: while predictive deltas benefit long-horizon tasks ($+2.4$ points on LIBERO-Long) and remain negligible on most routine manipulations, they in
Open source S3
benchmark success points
2.4
View original evidence
hifts average benchmark success ($+0.8$ points on LIBERO, $-0.8$ on RoboTwin~2.0), while generating visual futures increases inference latency by 1.7--1.9$\times$. These aggregate metrics conceal a concentrated failure mode that we term \emph{future-induced negative transfer}: while predictive deltas benefit long-horizon tasks ($+2.4$ points on LIBERO-Long) and remain negligible on most routine manipulations, they in
Open source S3

paper: not_reported

Source excerpts and review record

Automatically extracted; no manual editorial approval recorded.

hifts average benchmark success ($+0.8$ points on LIBERO, $-0.8$ on RoboTwin~2.0), while generating visual futures increases inference latency by 1.7--1.9$\times$. These aggregate metrics conceal a concentrated failure mode that we term \emph{future-induced negative transfer}: while predictive deltas benefit long-horizon tasks ($+2.4$ points on LIBERO-Long) and remain negligible on most routine manipulations, they in
Open source S3

duce severe degradation on precision tasks---with \emph{Press Stapler} and \emph{Place Can Basket} losing 54.0 and 34.0 points, recurring as top failure modes across independently trained WAM variants. We introduce \textbf{Utility-WAM} for in-place correction of future-conditioned control. Its primary learning component is a residual adapter that repairs future-induced velocity deltas under a frozen shared WAM by ali
Open source S4

Source：arXiv Robotics — research abstracts · arxiv.org

#Research #Reinforcement learning #World models #Sim-to-real #Evaluation