💡 [REQUEST] - Is there any guidance on how to reproduce the benchmark results?

[lirongxu-g]

起始日期 | Start Date

No response

实现PR | Implementation PR

No response

相关Issues | Reference Issues

No response

摘要 | Summary

Is there any guidance on how to reproduce the benchmark results?

基本示例 | Basic Example

for example, some python script

缺陷 | Drawbacks

none

未解决问题 | Unresolved questions

No response

[floe20]

Hi, please refer to the section about minicpm-o in the VLMEvalKit toolkit (vlmeval/vlm/minicpm_v.py). For most benchmarks, such as those in the OpenCompass leaderboard, we provide standard evaluation settings that you could follow.