You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello, thank you for this great work on Pass@k Training!
Most of the performance curves in the paper are based on Maze tasks. While these are helpful for controlled analysis, it would be very valuable to also see performance curves on math or code validation sets during training.
Would it be possible to share such curves to help verify whether the trends observed on Maze tasks also hold for more common reasoning benchmarks?