Meta LIMA Is Instruction Fine Tuning better than RLHF for LLM Alignment?

Meta LIMA, a 65B parameter LLaMa language model fine-tuned with the standard supervised loss on only 1000 carefully curated …

source