Reason--Imagine--Act: Closed-Loop LLM Decision Making with World Models for Autonomous Driving
The paper presents a new framework called Reason--Imagine--Act (RIA) for enhancing decision-making in autonomous driving using large language models (LLMs). This closed-loop system integrates an LLM reasoner with a world model to ensure safety during dynamic traffic scenarios. RIA demonstrates improved performance metrics compared to existing methods, achieving significant route completion and low collision rates.
- ▪RIA couples an LLM reasoner with an action-conditioned world model for online safety verification.
- ▪The framework achieved 80.05% route completion and a 0.20% collision rate in testing.
- ▪RIA outperforms training-free baselines on core closed-loop metrics.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.24004 (cs) [Submitted on 19 May 2026] Title:Reason--Imagine--Act: Closed-Loop LLM Decision Making with World Models for Autonomous Driving Authors:Zhengqi Sun, Yiwen Sun, Boxuan Liu, Tailai Chen, Tianxu Guo, Jiabin Liu View a PDF of the paper titled Reason--Imagine--Act: Closed-Loop LLM Decision Making with World Models for Autonomous Driving, by Zhengqi Sun and 5 other authors View PDF HTML (experimental) Abstract:Large language models (LLMs) are promising for autonomous driving, but semantics-only decision policies can yield physically unsafe behavior in dynamic traffic.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
