MMVU

The MMVU dataset/benchmark contains multiple-choice visual question-answering tasks and is used to evaluate the consistency between reasoning steps and final answers produced by multimodal large language models.

Papers using MMVU (3)

ProcessThinker: Enhancing Multi-modal Large Language Models Reasoning via Rollout-based Process Reward2026Enhancing Video Representations with Spatiotemporal-Semantic Residual to Mitigate Hallucinations in Video Large Multimodal Models2026Answer-consistent Chain-of-thought Reinforcement Learning For Multi-modal Large Langauge Models2025