What Do They See? Interpreting Complex Road Scenarios Through the Eyes of Vision-Language-Action Models for Safe and Trustworthy Autonomous Vehicle Learning

2026-07-18

AI Summary

This work addresses the opacity of end-to-end autonomous driving models in complex scenarios, which hinders safety assessment. To this end, the authors propose CVAA, a counterfactual ablation analysis framework that uniquely integrates generative image inpainting with mechanistic interpretability. By removing individual objects from front-view images to generate counterfactual samples, they construct the Counter-nuScenes dataset and analyze corresponding changes in model intermediate representations. Experiments on Alpamayo-1 and nuScenes demonstrate that CVAA can reveal whether the model’s causal dependencies on vehicles and pedestrians align with human expectations, while also uncovering anomalous responses to ostensibly irrelevant objects. This approach offers a novel pathway toward interpretable and trustworthy autonomous driving systems.

Abstract

End-to-end autonomous driving models are now able to navigate complex road scenarios, mapping raw sensor observations directly to observed paths for open-loop evaluation and often effective driving in closed-loop evaluation. Yet the internal logic of these safety-critical systems remains largely opaque, due to the complexity of traffic scenes. We propose a counterfactual ablation framework called Counterfactual Vision Action Analysis (CVAA) that systematically removes individual detected objects from front-camera images using photorealistic generative inpainting to prepare counterfactual sets to evaluate the difference in the model's response. This isolates the causal effect of each object's presence on the model's planning behaviour. Applied to the Alpamayo 1 trajectory predictor across 210 nuScenes driving scenes, we create a dataset Counter -nuScenes, using which we see that vehicles and pedestrians within the model's 'path' dominate causal influence as expected, while traffic lights, as expected, exert disproportionate effect relative to their image footprint. However, we also find cases where the model responds strongly to objects a human driver would consider irrelevant. This brings forth a deeper question: does the model itself view the scene as a sum of individual objects influencing the outcome, or does it encode an entirely different set of internal features that do not correspond to human-legible scene elements? To further understand this, we compare intermediate representations of original and inpainted image pairs using mechanistic interpretability techniques and examine the effect of the removal through the various model layers. Together, these two stages offer a path from behavioral auditing to representational understanding, creating explainable driving systems and solidifying human-AI trust.

Problem

• autonomous driving
• model interpretability
• vision-language-action models
• causal influence
• trustworthy AI

Innovation

• counterfactual ablation
• vision-language-action models
• mechanistic interpretability
• autonomous driving
• causal influence