ValsAI launches a leaderboard for SWE-bench Multimodal

A post says ValsAI launched a leaderboard for SWE-bench Multimodal, an updated SWE-bench variant focused on visual bugs. The poster cites maps rendering incorrectly and buttons appearing on the wrong side, and says there’s still lots to do even on this relatively simpler benchmark. In a follow-up reply, the poster credits @jyangballin and @closji with leading the project and says recent work helped enable the benchmark’s open-source launch.