Claude Code-Driving Scenario Mining for the Argoverse 2 Challenge
For autonomous driving researchers, this is an incremental application of existing LLM-based code generation to a specific benchmark challenge.
The authors present a pipeline for scenario mining in autonomous driving that uses a Claude Code agent for code generation, achieving submission to the CVPR 2026 Argoverse 2 Challenge. No concrete test set numbers are reported.
We present our submission to the CVPR 2026 Argoverse 2 Scenario Mining Challenge. Our system uses a four-stage pipeline: (1) autonomous code generation via a Claude Code agent powered by GLM~5.1, (2) iterative training set screening with Timestamp Balanced Accuracy threshold 0.8 to curate few-shot examples, (3) semantic code review by a separate Claude Code session, and (4) Qwen3-VL scene-level verification to filter false positives. We report results on the Argoverse 2 test set.