Submissions from github.com/swe-bench

		Top model scores may be skewed by Git history leaks in SWE-bench (github.com/swe-bench)
		466 points by mustaphah 7 months ago \| past \| 153 comments
		SWE-bench verified agents may look at future repository state (github.com/swe-bench)
		4 points by brrrrrm 7 months ago \| past