Tri-Info: Generalizable, Interpretable Failure Prediction for VLA Models via Information Theory

Across six VLA models and three benchmark environments, Tri-Info matches the strongest baselines in-domain. It comes from a paper posted to arXiv on 1 October 2026. Vision-Language-Action (VLA) models are increasingly deployed across diverse tasks, yet they remain black boxes whose physical interactions can cause irreversible harm, making generalizable and interpretable failure detection essential. We observe that successful and failed rollouts carry systematically different information-theoretic signatures. Building on this, we formalize VLA control as a closed-loop information pipeline and derive the Triple Information-theoretic (Tri-Info) signals that capture whether actions remain diverse, temporally consistent, and coupled to state transitions. Moreover, Tri-Info transfers across architectures, environments, and the sim-to-real gap without retraining with labeled data, reaching 70\% accuracy on real-world tasks. This establishes Tri-Info as a simple yet powerful method that not only detects failures with strong cross-domain generalization, but also delivers interpretable diagnostics of the underlying failure modes.

Why it counts

Across six VLA models and three benchmark environments, Tri-Info matches the strongest baselines in-domain.

Sources

Primary source: primary source

What is not known

This brief does not claim independent replication. Claims that appear only on X and not in the primary source stay unknown.