Need Conversion Script for ByteDance-Seed/Multi-SWE-bench to JSON Format Compatibility

[YANGCHEN205]

Hello,

I'm currently running Magentless experiments and noticed that while the default dataset (Daoguang/Multi-SWE-bench) provides Java-specific JSON data, the ByteDance-Seed/Multi-SWE-bench version contains multi-language support with different field structures.

Problem:

• Field mismatches between the two dataset formats
• Need JSON conversion for compatibility with existing tooling

Request:

1. Is there an existing script to convert ByteDance-Seed's multi-language format to the standard JSON structure?
2. If not, could documentation be provided about the field mapping requirements?

This would help maintain experiment consistency while leveraging the multi-language support.

Thank you!