# Teach-to-Crash: A Closed-Loop Student-Teacher LLM Framework for Collision-Inducing Test Scenario Generation

arXiv: https://arxiv.org/abs/2609.27296 · HTML: https://mlsanity.com/paper/2609.27296/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 14 of its 43 references that are in the map.
- First posted on arXiv 23 Sep 2026.

## Builds on (most cited first, 14 of 14)

- [Guided Conditional Diffusion for Controllable Traffic Simulation](https://mlsanity.com/paper/2210.17366/) (2022)
- [Generating Useful Accident-Prone Driving Scenarios via a Learned Traffic Prior](https://mlsanity.com/paper/2112.05077/) (2021)
- [ChatScene: Knowledge-Enabled Safety-Critical Scenario Generation for Autonomous Vehicles](https://mlsanity.com/paper/2405.14062/) (2024)
- [Scenic: A Language for Scenario Specification and Scene Generation](https://mlsanity.com/paper/1809.09310/) (2018)
- [Adversarial Evaluation of Autonomous Vehicles in Lane-Change Scenarios](https://mlsanity.com/paper/2004.06531/) (2020)
- [Failure-Scenario Maker for Rule-Based Agent using Multi-agent Adversarial Reinforcement Learning and its Application to Autonomous Driving](https://mlsanity.com/paper/1903.10654/) (2019)
- [ChatSUMO: Large Language Model for Automating Traffic Scenario Generation in Simulation of Urban MObility](https://mlsanity.com/paper/2409.09040/) (2024)
- [Compositional Falsification of Cyber-Physical Systems with Machine Learning Components](https://mlsanity.com/paper/1703.00978/) (2017)
- [Training Adversarial Agents to Exploit Weaknesses in Deep Control Policies](https://mlsanity.com/paper/2002.12078/) (2020)
- [Revisiting OPRO: The Limitations of Small-Scale LLMs as Optimizers](https://mlsanity.com/paper/2405.10276/) (2024)
- [TARGET: Automated Scenario Generation from Traffic Rules for Testing Autonomous Vehicles via Validated LLM-Guided Knowledge Extraction](https://mlsanity.com/paper/2305.06018/) (2023)
- [Reinforcement Learning Informed Evolutionary Search for Autonomous Systems Testing](https://mlsanity.com/paper/2308.12762/) (2023)
- [PAFOT: A Position-Based Approach for Finding Optimal Tests of Autonomous Vehicles](https://mlsanity.com/paper/2405.03326/) (2024)
- [Grammar-Constrained Refinement of Safety Operational Rules Using Language in the Loop: What Could Go Wrong](https://mlsanity.com/paper/2604.23523/) (2026)

## Abstract (arXiv)

Validating Autonomous Driving Systems (ADS) in simulation requires testing architectures that can discover rare, safety-critical failures while generating scenarios that are executable, diverse, and useful for downstream failure analysis. We introduce Teach-to-Crash, a closed-loop testing framework that combines a constrained ego-centric scenario representation, stagnation-aware search control, and a dual-LLM architecture for adaptive failure discovery. A high-reasoning Teacher LLM acts as an adaptive search controller, while a low-reasoning Student LLM emits simulator-executable scenarios in a strict JSON schema. The Teacher intervenes only when rolling collision rate and time-to-collision metrics stagnate, providing strategic guidance to redirect the search. In a CARLA case study with two experimental setups that vary the ego vehicle's speed policy, Teach-to-Crash achieves the highest Collision Hit Rate (90.79%), the shortest mean Time-to-Collision (18.31 s), and a competitive Collision Discovery Rate (136.21). PAFOT attains a higher mean CDR (179.44), but with substantially larger variance. Teach-to-Crash also yields the highest diversity (0.547) and, averaged across both setups on the CARLA Traffic Manager controller, the highest avoidability-based usefulness proxy (60.04%) among the compared methods. These results, within the evaluated CARLA scope, provide evidence that closed-loop dual-LLM reasoning can steer adversarial simulation-based testing over a constrained executable program space, generating failures that are frequent, structurally diverse, and assessed as more frequently avoidable.