MCTS result on MATH

[QQW-ing]

System Info

Operating System=Linux
Python version= 3.10
Hardware=A100 40g

Who can help?

@ziyuwan

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the codebase (such as scrips/, ...)
• My own task or dataset (give details below)

Reproduction

Hi, thanks for your excellent work.

When I use "Qwen2.5-Math-7B-Instruct" as the base model, I got "[{"majority_vote": 0.824, "total_completion_tokens": 2990.498}]". However, when I use another base model like 'mistral-7b-sft' on MATH , I got result '[{"majority_vote": 0.29, "total_completion_tokens": 963.846}]' in vanilla MCTS process.

Expected behavior

I wonder if the result is reasonable. And can you give some results based on your experiment? Many thanks!!!