chujiezheng/Smaug-Llama-3-70B-Instruct-ExPO

Smaug-Llama-3-70B-Instruct-ExPO is a 70 billion parameter instruction-tuned language model developed by chujiezheng, based on abacusai/Smaug-Llama-3-70B-Instruct and Meta-Llama-3-70B-Instruct. This model utilizes an extrapolation (ExPO) method with an alpha of 0.3 to enhance alignment with human preferences. It demonstrates improved win rates on the AlpacaEval 2.0 benchmark and higher scores on MT-Bench compared to its base models, making it suitable for applications requiring strong conversational and instruction-following capabilities.