[Suggestion] Evaluate Ming on the WorldSense Benchmark

[shilinyan99]

Hi Ming Team,

First off, congratulations on the excellent work on Ming! We've been following your project and are truly impressed by its powerful audio-video co-understanding capabilities. Your contribution to the multimodal community is fantastic.

We are the maintainers of WorldSense, a benchmark specifically designed to evaluate a model's ability in real-world audio-video understanding. Given Ming's advanced multimodal performance, we believe it would be an excellent candidate to test on our benchmark. An evaluation on WorldSense could effectively highlight your model's unique strengths from a new and valuable perspective.

Here is some information about our benchmark:

Paper: https://arxiv.org/pdf/2502.04326
GitHub: https://github.com/JaaackHongggg/WorldSense
Leaderboard: https://jaaackhongggg.github.io/WorldSense/#leaderboard

We would be thrilled to see how Ming performs and would like to kindly invite you to evaluate it on the WorldSense benchmark. A strong performance would further showcase its leading capabilities to the community.

We would be more than happy to provide any support needed during the evaluation, such as clarifying data formats or assisting with the evaluation script.

Thank you for your time and for your incredible work!

Best regards,

Shilin Yan

[BaldrLector]

Dear Shilin Yan,

Thank you so much for reaching out and for your kind words about Ming! We are truly honored to hear that you’ve been following our work and appreciate your encouragement.

We’re also very excited to learn about WorldSense—it looks like a thoughtfully designed benchmark that addresses a critical aspect of multimodal understanding. Evaluating Ming on WorldSense would indeed provide valuable insights into its real-world audio-video comprehension capabilities, and we’re keen to explore this opportunity.

We’ve reviewed the resources you shared (the paper, GitHub repository, and leaderboard) and are impressed with the scope and rigor of the benchmark. Our team will begin looking into the evaluation process and data requirements. Should we have any questions or need assistance, we’ll certainly reach out—thank you for your generous offer to support us during the evaluation.

Once again, thank you for your invitation and for the impactful work you’re doing with WorldSense!

Best regards,

Guangming Yao
Ming Team