Search papers, labs, and topics across Lattice.
This paper addresses the challenge of modality entanglement in Vision-Language-Action (VLA) policies, which leads to spurious correlations and poor performance in real-world scenarios. The authors introduce Evidence-Gated Regularization (EGR), a modality-agnostic training objective that enhances task-relevant signal extraction without adding inference-time overhead. EGR significantly improves success rates across various benchmarks and real-robot setups, demonstrating its effectiveness in mitigating the effects of uninformative sensors and enhancing single-sensor performance.
EGR boosts robot performance by up to 183% in challenging environments by effectively distinguishing between informative and uninformative sensory inputs.
Vision-Language-Action (VLA) policies fuse multimodal sensory inputs, but training on limited and homogeneous robot demonstrations encourages spurious inter-sensor correlations rather than task-relevant signal, a failure we term modality entanglement. Under real-world occlusions and distractors, this manifests as nuisance sensitivity to corruption of uninformative sensors and single-modality insufficiency when only one informative sensor remains intact. We propose Evidence-Gated Regularization (EGR), a modality-agnostic training objective that introduces zero inference-time overhead. EGR derives a per-frame and per-sensor task-relevance signal to gate two state-conditional consistency objectives: invariance on low-evidence sensors, and single-sensor sufficiency on high-evidence ones. We introduce a benchmark based on BEHAVIOR-1K, comprising a fast inference-only diagnostic suite and 47 rollout-based skills targeting modality entanglement. We validate EGR on this benchmark and on two real-robot setups with fundamentally different embodiments: a bi-manual setup with two Kinova arms and three RGB cameras, and a single-arm MELFA ASSISTA setup combining vision and GelSight tactile sensors. EGR improves simulation success rates (SR) from 12.5% to 16.4% under full modalities (+31%), from 9.4% to 16.5% under uninformative-sensor corruption (+75%), and from 2.8% to 6.1% under single-sensor fallback (+120%). Under physical-object distractors, EGR boosts SR from 30% to 85% on the bi-manual setup (+183%) and from 55% to 70% on the tactile setup (+27%).
