DRIV-EX: Counterfactual Explanations for Driving LLMs

2026-02-28

AI Summary

This work addresses the limited interpretability of large language models (LLMs) in autonomous driving decision-making, where it is often difficult to trace the reasoning behind their predictions. To enhance transparency, the authors propose a counterfactual explanation method that leverages gradient-based optimization over continuous embeddings to guide minimal semantic modifications of input scenarios, combined with controlled decoding to generate fluent and contextually faithful natural language descriptions of these changes. The approach significantly improves the transparency and robustness of an LLM-driven planner (LC-LLM). Experimental results on textual transcriptions from the highD dataset demonstrate that the proposed method reliably produces higher-quality and more valid counterfactual explanations compared to baseline approaches, effectively uncovering potential model biases and offering a promising direction toward explainable autonomous driving systems.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Interpretability, Explainability, and Transparency
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Large language models (LLMs) are increasingly used as reasoning engines in autonomous driving, yet their decision-making remains opaque. We propose to study their decision process through counterfactual explanations, which identify the minimal semantic changes to a scene description required to alter a driving plan. We introduce DRIV-EX, a method that leverages gradient-based optimization on continuous embeddings to identify the input shifts required to flip the model's decision. Crucially, to avoid the incoherent text typical of unconstrained continuous optimization, DRIV-EX uses these optimized embeddings solely as a semantic guide: they are used to bias a controlled decoding process that re-generates the original scene description. This approach effectively steers the generation toward the counterfactual target while guaranteeing the linguistic fluency, domain validity, and proximity to the original input, essential for interpretability. Evaluated using the LC-LLM planner on a textual transcription of the highD dataset, DRIV-EX generates valid, fluent counterfactuals more reliably than existing baselines. It successfully exposes latent biases and provides concrete insights to improve the robustness of LLM-based driving agents.

Problem

• large language models
• autonomous driving
• counterfactual explanations
• decision interpretability
• opaque decision-making

Innovation

• counterfactual explanations
• large language models
• autonomous driving
• gradient-based optimization
• controlled decoding