RuCL: Stratified Rubric-Based Curriculum Learning for Multimodal Large Language Model Reasoning
This addresses inefficient training dynamics in rubric-based approaches for multimodal reasoning, though it appears incremental as it builds on existing RLVR and rubric-based methods.
The paper tackles the problem of reward hacking in multimodal large language model reasoning by proposing RuCL, a stratified rubric-based curriculum learning framework that dynamically adjusts rubric weights during training, resulting in a +7.83% average improvement over the Qwen2.5-VL-7B model and achieving state-of-the-art accuracy of 60.06%.
Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a prevailing paradigm for enhancing reasoning in Multimodal Large Language Models (MLLMs). However, relying solely on outcome supervision risks reward hacking, where models learn spurious reasoning patterns to satisfy final answer checks. While recent rubric-based approaches offer fine-grained supervision signals, they suffer from high computational costs of instance-level generation and inefficient training dynamics caused by treating all rubrics as equally learnable. In this paper, we propose Stratified Rubric-based Curriculum Learning (RuCL), a novel framework that reformulates curriculum learning by shifting the focus from data selection to reward design. RuCL generates generalized rubrics for broad applicability and stratifies them based on the model's competence. By dynamically adjusting rubric weights during training, RuCL guides the model from mastering foundational perception to tackling advanced logical reasoning. Extensive experiments on various visual reasoning benchmarks show that RuCL yields a remarkable +7.83% average improvement over the Qwen2.5-VL-7B model, achieving a state-of-the-art accuracy of 60.06%.