Hacker News new | ask | show | jobs
by peakji 603 days ago
It is an LLM fine-tuned using a new type of dataset and RL reward. It's good at reasoning, but I would not recommend to replace Llama for general tasks.