Research papers · Posted 2026.02.24 · Preprint
An LLM-driven Scenario Generation Pipeline Using an Extended Scenic DSL for Autonomous Driving Safety Validation
An LLM-driven Scenario Generation Pipeline Using an Extended Scenic DSL for Autonomous Driving Safety Validation
Overview
The pipeline converts crash-report text and sketches into an extended Scenic intermediate representation using an LLM. CARLA and Autoware experiments test semantic extraction and reproduction of intended traffic-rule violations across scenario variations.
Version: v1