We found 110 reward hacks and grading bugs in Terminal-Bench, DeepSWE, and BFCL (usetokenless.com)

2 points by random17 20d ago ↗ HN

1 comment

[ 65.4 ms ] story [ 678 ms ] thread

rohaga 20d ago ↗

hey everyone, I’m the author of this blog post, happy to answer any questions in this thread!