Announcement
ProgramBench opens submissions for model and harness runs
A post links to Muse Spark 1.3 evaluation profiles and says trajectories, final codebases and test pass/fail breakdowns have been open-sourced.
TLDR
A post says ProgramBench submissions are open for people to run their own model or harness. ProgramBench’s guide describes how to turn an evaluation run into a public leaderboard submission. The poster also says Muse Spark 1.3 trajectories, final codebases and test pass/fail breakdowns are open source, and describes ProgramBench as a joint effort across Meta Superintelligence Labs, Stanford and Harvard.
Combined views
—
1 Source, first seen ago
