Teach-to-Crash: A Closed-Loop Student-Teacher LLM Framework for Collision-Inducing Test Scenario Generation
Source
Zaid Ghazal, Khouloud Gaaloul, Bruce Maxim
Author
Zaid Ghazal, Khouloud Gaaloul, Bruce Maxim
Date
Key takeaways · AI-distilled
A low-reasoning Student LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition → emits simulator-executable scenarios in a strict JSON schema, while the Teacher steps in only when rolling collision rate and time-to-collision metrics stagnate.
Beyond hit rate, Teach-to-Crash had the shortest mean time-to-collision (18.31 s) and the highest scenario diversity (0.547) among compared methods in the CARLA study.
PAFOT found collisions at a higher mean discovery rate (179.44 vs 136.21), though with substantially larger variance.
The authors limit their claims to the evaluated CARLA scope; averaged over both setups on the CARLA Traffic Manager controller, it also scored highest on an avoidability-based usefulness proxy (60.04%).
Terms in this piece · Glossary
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
Why it matters
Offers a reusable pattern for using a high-reasoning LLM as a search controller that only intervenes when a lower-cost model's test generation stalls.
articleATLAS: Discovering Agent Strategies through LLM-Guided Abstraction and Automata LearningIgnacio D. Lopez-Miguel, Andreas Happe, J\"urgen Cito, Ezio Bartocci, Bettina K\"onighofer, Martin Tappler
articleA Closed-Loop Control Architecture for Reliable Constraint Satisfaction in LLM Text GenerationQuan Zhou, Shahbaz Siddeeq, Mika Saari, Pekka Abrahamsson
articletrajectory-judge: What Outcome-Only LLM Judges Miss on Agent TrajectoriesHadi Mohammadi