Search papers, labs, and topics across Lattice.
This paper introduces a framework for evaluating and improving calibration in vision-language-action (VLA) models operating in sequential tasks with partial trajectory observations. They formulate a sequential extension of the Brier score, demonstrating that its risk minimizer aligns with the VLA policy's value function, effectively linking uncertainty calibration and reinforcement learning. Empirically, they show that Temporal Difference (TD) calibration enhances performance on both simulated and real-robot data, and surprisingly, that TD-calibrated single-step action probabilities can yield competitive uncertainty estimates.
Reinforcement learning's Temporal Difference value estimation offers a surprisingly effective and theoretically grounded approach to calibrating uncertainty in vision-language-action models for robotics.
Recent advances in vision-language-action (VLA) models for robotics have highlighted the importance of reliable uncertainty quantification in sequential tasks. However, assessing and improving calibration in such settings remains mostly unexplored, especially when only partial trajectories are observed. In this work, we formulate sequential calibration for episodic tasks, where task-success confidence is produced along an episode, while success is determined at the end of it. We introduce a sequential extension of the Brier score and show that, for binary outcomes, its risk minimizer coincides with the VLA policy's value function. This connection bridges uncertainty calibration and reinforcement learning, enabling the use of temporal-difference (TD) value estimation as a principled calibration mechanism over time. We empirically show that TD calibration improves performance relative to the state-of-the-art on simulated and real-robot data. Interestingly, we show that when calibrated using TD, the VLA's single-step action probabilities can yield competitive uncertainty estimates, in contrast to recent findings that employed different calibration techniques.
