--- license: mit library_name: transformers pipeline_tag: video-text-to-text --- # TAU-R1 This repository contains the model presented in [TAU-R1: Visual Language Model for Traffic Anomaly Understanding](https://huggingface.co/papers/2603.19098). TAU-R1 is a two-layer vision-language framework for Traffic Anomaly Understanding (TAU). The first layer is a lightweight anomaly classifier that performs coarse anomaly categorisation, while the second layer is a larger anomaly reasoner that generates detailed event summaries. It is trained with a two-stage strategy consisting of decomposed-QA-enhanced supervised fine-tuning followed by TAU-GRPO, a GRPO-based post-training method with TAU-specific reward functions. Code: https://github.com/siri-rouser/TAU-R1.