Alpaca-7B-4bit26.66%44.32%0.00%0.00% Alpaca-7B-4bit cleaned37.00%--12.90% The cleaned dataset appears to hallucinate less and perform better than the original dataset. Alpaca is a fine-tuned version of LLAMA tha
Alpaca-7B-4bit26.66%44.32%0.00%0.00% Alpaca-7B-4bit cleaned37.00%--12.90% The cleaned dataset appears to hallucinate less and perform better than the original dataset. Alpaca is a fine-tuned version of LLAMA that was trained using an Instruct Dataset generated by GPT-3. The generated dataset...