RubricHub: A Comprehensive and Highly Discriminative Rubric Dataset via Automated Coarse-to-Fine Generation

TL;DR

RubricHub introduces a fully automated Coarse-to-Fine Rubric Generation framework to overcome supervision ceilings in open-ended generation. By combining response grounding, principle guidance, multi-model aggregation, and difficulty evolution, it creates ~110k cross-domain rubrics with high discriminability. When integrated into a two-stage post-training pipeline (RuFT and RuRL), Qwen3-14B achieves state-of-the-art HealthBench performance and surpasses GPT-5 in this domain, demonstrating the practical value of fine-grained rubrics for alignment. The work also provides insights into rubric reliability, efficiency, and the trade-offs involved in scalable rubric-driven RL for non-ground-truth tasks.

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) has driven substantial progress in reasoning-intensive domains like mathematics. However, optimizing open-ended generation remains challenging due to the lack of ground truth. While rubric-based evaluation offers a structured proxy for verification, existing methods suffer from scalability bottlenecks and coarse criteria, resulting in a supervision ceiling effect. To address this, we propose an automated Coarse-to-Fine Rubric Generation framework. By synergizing principle-guided synthesis, multi-model aggregation, and difficulty evolution, our approach produces comprehensive and highly discriminative criteria capable of capturing the subtle nuances. Based on this framework, we introduce RubricHub, a large-scale ($\sim$110k) and multi-domain dataset. We validate its utility through a two-stage post-training pipeline comprising Rubric-based Rejection Sampling Fine-Tuning (RuFT) and Reinforcement Learning (RuRL). Experimental results demonstrate that RubricHub unlocks significant performance gains: our post-trained Qwen3-14B achieves state-of-the-art (SOTA) results on HealthBench (69.3), surpassing proprietary frontier models such as GPT-5. The code and data will be released soon.

Figures (9)

• Figure 1: Motivating Example. Comparison between coarse-grained and fine-grained evaluation. Coarse rubrics (Rubric 1) result in indistinguishable high scores, whereas RubricHub (Rubric 2) utilizes highly discriminative criteria to reveal specific weaknesses, providing richer signals for alignment.
• Figure 2: Overall method pipeline. (a) Coarse-to-Fine Rubric Generation: Candidates are synthesized via response-grounded and principle-guided strategies, then refined through aggregation and difficulty evolution into RubricHub. (b) Utilization of Rubric in Post-Training : Rubrics are applied in RuFT (left) for rejection sampling and in RuRL (right) to provide structured reward signals for policy optimization.
• Figure 3: Pie chart showing the source distribution across five major domains.
• Figure 4: Score density distribution across models.
• Figure 5: Performance comparison using RaR and RubricHub in Medical (left) and Science (right) domains on Qwen3-14B-Base. RaR (original): original RaR dataset. RaR (Rubrics by RubricHub): RaR questions with Rubrics regenerated by our pipeline.