Search papers, labs, and topics across Lattice.
The paper introduces ICR-Drive, a diagnostic benchmark to evaluate the instruction-following robustness of language-conditioned autonomous driving agents. ICR-Drive generates counterfactual instruction variants across paraphrase, ambiguity, noise, and misleading categories and measures the resulting performance degradation on fixed CARLA routes. Experiments with LMDrive and BEVDriver show significant performance drops with minor instruction changes, highlighting a robustness gap in current models.
Even slight variations in natural language instructions can cause language-driven autonomous driving models to fail dramatically, revealing a critical reliability gap for real-world deployment.
Recent progress in vision-language-action (VLA) models has enabled language-conditioned driving agents to execute natural-language navigation commands in closed-loop simulation, yet standard evaluations largely assume instructions are precise and well-formed. In deployment, instructions vary in phrasing and specificity, may omit critical qualifiers, and can occasionally include misleading, authority-framed text, leaving instruction-level robustness under-measured. We introduce ICR-Drive, a diagnostic framework for instruction counterfactual robustness in end-to-end language-conditioned autonomous driving. ICR-Drive generates controlled instruction variants spanning four perturbation families: Paraphrase, Ambiguity, Noise, and Misleading, where Misleading variants conflict with the navigation goal and attempt to override intent. We replay identical CARLA routes under matched simulator configurations and seeds to isolate performance changes attributable to instruction language. Robustness is quantified using standard CARLA Leaderboard metrics and per-family performance degradation relative to the baseline instruction. Experiments on LMDrive and BEVDriver show that minor instruction changes can induce substantial performance drops and distinct failure modes, revealing a reliability gap for deploying embodied foundation models in safety-critical driving.
