Popular repositories Loading
-
tau-bench-judge-audit
tau-bench-judge-audit PublicDoes τ-bench's hardcoded GPT-4.1 judge favour its own model family? A pre-registered re-grading study — 320 trajectories, 1,792 judge calls, $13.76. No bias detectable at MDE 0.137; a 9.1-pt judge-…
Python
Something went wrong, please refresh the page to try again.
If the problem persists, check the GitHub status page or contact support.
If the problem persists, check the GitHub status page or contact support.