3 ms·
Reinforcement Learning for Reasoning in LLMs with One Training Example
- krackers 1y agoTurns out that https://www.fortressofdoors.com/four-magic-words/ https://www.fortressofdoors.com/four-magic-words/ was right and all you need to do in training is have the LLM meditate on a single example.