We-Math
Emerging4papers using it
2025first seen
The 'We-Math' dataset is used to evaluate the reasoning ability of multi-modal large language models (MLLMs) in mathematical problem-solving without requiring ground truth labels.
Papers using We-Math (4)
- Advancing Multimodal Reasoning: From Optimized Cold Start to Staged Reinforcement LearningAthena: Enhancing Multimodal Reasoning with Data-efficient Process Reward ModelsAdvancing Multimodal Reasoning via Reinforcement Learning with Cold StartFirst SFT, Second RL, Third UPT: Continual Improving Multi-Modal LLM Reasoning via Unsupervised Post-Training