Blog
Updates and announcements from the PostTrainBench team.
PostTrainBench v1.2
BFCL removed, scoring fixes, multi-seed evaluation, majority vote judging, and Harbor support.
Our response to Epoch AI's review of PostTrainBench
Where we agree with Epoch AI's review of v1.1, and what it changes for v1.2.
PostTrainBench v1.1
A more precise ruleset and an improved anti-cheat pipeline.