What Do They See? Interpreting Complex Road Scenarios Through the Eyes of Vision-Language-Action Models for Safe and Trustworthy Autonomous Vehicle Learning
A counterfactual ablation framework called Counterfactual Vision Action Analysis (CVAA) is proposed 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.
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.
In autonomous driving, Bird's-Eye View (BEV) representations provide a structured, top-down abstraction of the vehicle's surroundings and have become a key input modality for Behavioral Cloning (BC) policies. While ground-truth BEV maps are readily available in simulation, real-world deployment requires replacing them...
Felipe Dos Santos, Eric A. Antonelo, Gustavo Claudio Karl Couto· 0 citations
Visual Evidence Augmentation, a recent inference-time method, is applied to examine whether it can improve model reliability under degraded visual conditions and finds that it improves performance for some models and datasets, although the gains are not consistent across all settings.
Manasa Mariam Mammen, Priyanka Mary Mammen, Zafer Kayatas et al.· 0 citations
This work uses intervention-based metrics that measure how visual observations, instructions, and visual memory causally influence navigation decisions and shows that these navigation policies are sensitive to all input modalities and do not depend on a single one.
Geo-VLA is proposed, a plug-and-play framework that enhances VLA models by learning geometry-aware visual representations and introduces Geo-QA, a geometry-focused question-answering dataset that injects road geometry into vision-language representations through contrastive learning and instruction tuning.
Ran Chen, Jiaxing Ren, Zhikun Zhang et al.· 0 citations
Retrieval-augmented generation (RAG) gives vision--language driving systems access to external safety knowledge, yet a retrieved risk rule may be relevant without applying to the current scene. A vision--language model (VLM) receiving such knowledge must ground objects, bind entities across time, and verify relations b...