do we hava a plan feature about Rft?

[zjrwtx]

Feature request

https://openai.com/form/rft-research-program/

Motivation

Enhance model professionalism: Reinforcement Fine-Tuning (RFT) is designed to optimize models through high-quality task data sets to make the model perform more accurately in complex tasks in a specific domain, thereby elevating the model from "high school level" to "doctoral level expert" capability.

Reduced training data requirements: RFT technology allows developers to fine-tune models using data sets from tens to thousands of high-quality tasks, meaning significant performance gains can be achieved even with limited data (sometimes just a few dozen samples).

Enhanced reasoning: Unlike traditional fine-tuning, reinforcement fine-tuning does not simply make the model "remember the answer", but by training the model to learn to reason in a specific domain, to find the right answer, thereby improving the model's ability to solve similar problems.

Your contribution

i guess i can cooperate with you guys

[YanSong97]

Yes absolutely! We add RFT to our TODO list with high priority! We welcome any form of collaboration!

[zjrwtx]

got it,thanks!

[kargarisaac]

Is anyone working on this? I'm interested to contribute.