RuCL: Stratified Rubric-Based Curriculum Learning for Multimodal Large Language Model Reasoning

TL;DR

RuCL introduces a curriculum learning framework for multimodal large language models that uses stratified rubrics and dynamic reward weighting to improve reasoning capabilities efficiently.

Contribution

It proposes a novel stratified rubric-based curriculum learning method that enhances reasoning in multimodal LLMs by focusing on reward design and competence-based rubric stratification.

Findings

Achieves +7.83% average improvement over baseline models.

Reaches a state-of-the-art accuracy of 60.06% on visual reasoning benchmarks.

Demonstrates effective guidance from perception to logical reasoning.

Abstract

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%.