Uni-MMMU: A Massive Multi-discipline Multimodal Unified Benchmark
This provides a reliable foundation for advancing unified multimodal models, addressing a gap in evaluation for researchers and developers in AI.
The authors tackled the lack of benchmarks for truly integrated visual understanding and generation in unified multimodal models by introducing Uni-MMMU, a comprehensive benchmark across eight domains, revealing substantial performance disparities and cross-modal dependencies among state-of-the-art models.
Unified multimodal models aim to jointly enable visual understanding and generation, yet current benchmarks rarely examine their true integration. Existing evaluations either treat the two abilities in isolation or overlook tasks that inherently couple them. To address this gap, we present Uni-MMMU, a comprehensive and discipline-aware benchmark that systematically unfolds the bidirectional synergy between generation and understanding across eight reasoning-centric domains, including science, coding, mathematics, and puzzles. Each task is bidirectionally coupled, demanding models to (i) leverage conceptual understanding to guide precise visual synthesis, or (ii) utilize generation as a cognitive scaffold for analytical reasoning. Uni-MMMU incorporates verifiable intermediate reasoning steps, unique ground truths, and a reproducible scoring protocol for both textual and visual outputs. Through extensive evaluation of state-of-the-art unified, generation-only, and understanding-only models, we reveal substantial performance disparities and cross-modal dependencies, offering new insights into when and how these abilities reinforce one another, and establishing a reliable foundation for advancing unified models.