analiza aiqbench
This research demonstrates that frozen predictive states within robotic world models can be efficiently decoded to monitor execution failures without retraining the core backbone. By training a lightweight readout on these internal representations, the method achieves high accuracy in distinguishing failure trajectories while adding negligible latency to real-robot deployments. The findings validate the utility of pretrained world model states as reusable features for robust, causal risk assessment in autonomous systems.
FARM: Reading Failure Signals from the Internal Predictive States of a Frozen Robotic World Model
arXiv:2609.11445v1 Announce Type: new Abstract: Reliable robot deployment requires online failure monitoring, yet existing monitors mainly derive risk from proxy signals or train dedicated monitoring components. We ask whether the internal predictive states of a frozen pretrained robotic world model already contain directly decodable failure information. Failure-Aware Readout from World Models (FARM) trains only a 33,985-parameter supervised readout over frozen VLA-JEPA predictive states, producing step-wise failure scores and causal trajectory risk. Five-fold out-of-fold evaluation across seven source tasks reaches 85.68/88.59 pooled AUROC/AUPRC, and FARM gives the best Seen performance among 15 matched baselines on the 10-task benchmark. Across four real-robot populations on PIPER X, SO-101, and Franka, fixed-readout transfer and readout-only adaptation test deployment shifts without updating the predictive backbone. FARM also discriminates failures from partial causal histories and adds 0.2256 ms mean CUDA latency once the frozen state is available. These results support frozen predictive world-model states as reusable features for causal, transferable, and low-overhead execution monitoring.
Abstract:Reliable robot deployment requires online failure monitoring, yet existing monitors mainly derive risk from proxy signals or train dedicated monitoring components. We ask whether the internal predictive states of a frozen pretrained robotic world model already contain directly decodable failure information. Failure-Aware Readout from World Models (FARM) trains only a 33,985-parameter supervised readout over frozen VLA-JEPA predictive states, producing step-wise failure scores and causal trajectory risk. Five-fold out-of-fold evaluation across seven source tasks reaches 85.68/88.59 pooled AUROC/AUPRC, and FARM gives the best Seen performance among 15 matched baselines on the 10-task benchmark. Across four real-robot populations on PIPER X, SO-101, and Franka, fixed-readout transfer and readout-only adaptation test deployment shifts without updating the predictive backbone. FARM also discriminates failures from partial causal histories and adds 0.2256 ms mean CUDA latency once the frozen state is available. These results support frozen predictive world-model states as reusable features for causal, transferable, and low-overhead execution monitoring.
Robotics (cs.RO)
arXiv:2609.11445 [cs.RO]
(or arXiv:2609.11445v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.11445 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.
Text integral sub licența deschisă a publicației · 246 cuvinte
Subiecte conexe
The Entangled Dimensions of Decision Theory
LessWrong · 30 iul. 2026
Models Can't Remember Their Training. Neither Can You.
LessWrong · 19 iul. 2026
The current bottleneck is political will, not research
AI Alignment Forum · 11 iul. 2026
Game-Theoretic Disempowerment: Loss of Control to Agencyless AI
LessWrong · 27 sept. 2026