RubricHub is a large-scale, multi-domain generative task dataset released in 2026 by Li Auto in collaboration with Zhejiang University. Related research papers include... RubricHub: A Comprehensive and Highly Discriminative Rubric Dataset via Automated Coarse-to-Fine Generation This dataset provides high-quality supervision based on scoring criteria for open-ended generation tasks. The dataset is constructed using an automated coarse-to-fine scoring criterion generation framework, integrating strategies such as principle-guided synthesis, multi-model aggregation, and difficulty evolution to produce comprehensive and highly discriminative evaluation criteria.

Dataset composition:

• RuRL includes a list of prompts in the VeRL training format, data sources, capability objectives, reward model, and additional information.
• RuFT includes data sources, model input hints, model-generated responses, sample IDs, scoring criteria, and scoring details.