TrafficAlign: Aligning Large Language Models for Traffic Scenario Generation
About
Recent research has investigated the use of large language models (LLMs) to generate traffic scenarios for autonomous driving. However, pretrained LLMs often fail to align with real-world traffic distributions. In this work, we present TrafficAlign, an automated framework that synthesizes traffic scenarios based on real-world driving videos, performs data validation, and aligns LLMs with the synthesized scenarios. The evaluation shows that traffic scenarios generated by TrafficAlign are highly effective, revealing up to 10.8% more collisions on average across three autonomous driving models than state-of-the-art methods. Furthermore, fine-tuning these driving models with TrafficAlign-generated scenarios significantly reduced collision rates by 36.1% compared with the original models. A qualitative study using traffic datasets from six geographically diverse regions shows that TrafficAlign-generated scenarios exhibit strong alignment with corresponding traffic distributions in these regions.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Autonomous Driving Model Testing | CARLA SafeBench (Multiple Regions) | Compliance Ratio (CR)93.3 | 11 | |
| Autonomous Driving Evaluation | Composite Los Angeles (held-out set) | CR70.1 | 3 | |
| Autonomous Driving Evaluation | Composite (New York City) | Completion Rate (CR)71 | 3 | |
| Autonomous Driving Evaluation | Composite held-out set (Yellowstone) | CR74.3 | 3 | |
| Autonomous Driving Evaluation | Composite Yosemite (held-out set) | CR60.4 | 3 | |
| Autonomous Driving Evaluation | Composite Small Towns (held-out set) | CR62.9 | 3 | |
| Autonomous Driving Evaluation | Composite held-out set (Switzerland) | CR62.5 | 3 |