Chat2Scenic: An Iterative RAG-Based Framework for Scenario Generation in Autonomous Driving
Chat2Scenic reports a 76.42% script compilation success rate on a new 123-scenario benchmark.
The paper frames autonomous-driving test generation as executable DSL script creation from regulatory descriptions. Its system uses iterative RAG with a chatbot interface to refine scenarios against regulatory knowledge and DSL syntax. The benchmark draws from NHTSA, UN vehicle regulations, and other sources. The authors say it beats retrieval-assemble and full-script retrieval baselines on both compilation success and framework accuracy. HF Daily Papers' note
The paper frames autonomous-driving test generation as executable DSL script creation from regulatory descriptions. Its system uses iterative RAG with a chatbot interface to refine scenarios against regulatory knowledge and DSL syntax. The benchmark draws from NHTSA, UN vehicle regulations, and other sources. The authors say it beats retrieval-assemble and full-script retrieval baselines on both compilation success and framework accuracy. HF Daily Papers' note
score 4