Should I use CoT prompting in RL model as instruction-tuned model?
#3
by
tongyx361
- opened
Example in the RL model card does not contain CoT prompting.
Thanks for your question! We have updated the model card. We recommend using CoT prompting to obtain the best performance.
If you also want to activate tool-integrated reasoning, please check out the link below:
https://github.com/deepseek-ai/DeepSeek-Math/tree/main/evaluation
ZhihongShao
changed discussion status to
closed