Paper or Dataset Available?
#2
by
Broyojo
- opened
Is there a paper available or material available on how this PRM was trained? It would be very valuable to know for reproducibility and what use case this PRM is good for based on its training, i.e. is it trained using MathShepherd strategy, making it a useful Q value estimator, or is it trained using PRM800K style so it assesses step correctness instead?