Discussion on swebench-extra Modifications and Data Generation Issues
#2
by
Thomas-147
- opened
I've recently been trying to use swebench-extra to synthesize some trajectory data, but found that some of the data fails during 'git apply'. I noticed that the patch format in your dataset has been modified—may I ask what improvements you've made based on the original swebench-extra, and whether you encountered similar issues?
By the way, I'm also currently working on synthesizing swebench data, focusing on Python datasets (like swe-gym and swebench-extra) as well as data for other programming languages. Since you're working on similar projects, could you email me your contact information for further discussion? My email is [email protected].