MMEnterprise-Bench
A multimodal record-evolution benchmark for enterprise agents, built from simulated collaborative projects with 1,613 questions. It covers three task types—evidence reasoning, conflict resolution and unanswerability identification—and is used to test agents' reasoning over dynamic multimodal information and their ability to assess evidence sufficiency.
Timeline (last 90 days)
MMEnterprise-Bench is an evaluation benchmark for enterprise Agents that uses multimodal records in simulated collaborative projects to test Agents' reasoning capabilities.
MMEnterprise-Bench built 1,613 questions based on simulated collaborative projects, covering three types of tasks: evidence reasoning, conflict resolution, and unanswerability identification.
In the MMEnterprise-Bench experiments, the research team evaluated 10 closed-source and open-source multimodal large models.
The paper's experiments show that even the strongest models struggle in the face of cross-source or temporal conflicts.
The code and data for MMEnterprise-Bench have been made public.
All Facts (5)
MMEnterprise-Bench is an evaluation benchmark for enterprise Agents that uses multimodal records in simulated collaborative projects to test Agents' reasoning capabilities.
50%UnverifiedMMEnterprise-Bench built 1,613 questions based on simulated collaborative projects, covering three types of tasks: evidence reasoning, conflict resolution, and unanswerability identification.
50%UnverifiedIn the MMEnterprise-Bench experiments, the research team evaluated 10 closed-source and open-source multimodal large models.
50%UnverifiedThe paper's experiments show that even the strongest models struggle in the face of cross-source or temporal conflicts.
50%UnverifiedThe code and data for MMEnterprise-Bench have been made public.
50%