Brief, mostly facts
MMEnterprise-Bench tests enterprise agents on evolving multimodal records
An arXiv benchmark of 1,613 questions across 10 MLLMs, with code and data released. It covers evidence reasoning, conflicts and unanswerable cases.
What the research found
MMEnterprise-Bench is an arXiv benchmark of 1,613 questions over evolving multimodal workplace records, testing evidence reasoning, conflict resolution and unanswerability across 10 MLLMs, with code and data released.
Claims and how well the evidence supports them
- MMEnterprise-Bench has 1,613 questions 99%
- Evaluates 10 proprietary and open-source multimodal LLMs 99%
- Code and data are available at mlvlab.github.io/MMEnterprise-Bench 97%
Evidence
- huggingface.co https://huggingface.co/papers/2610.27911
- mlvlab.github.io https://mlvlab.github.io/MMEnterprise-Bench/
Actionability rated medium.