chujiezheng/Starling-LM-7B-alpha-ExPO
TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Apr 26, 2024License:apache-2.0Architecture:Transformer Open Weights Warm

The chujiezheng/Starling-LM-7B-alpha-ExPO is a 7 billion parameter language model, extrapolated (alpha = 0.2) from berkeley-nest/Starling-LM-7B-alpha and openchat/openchat_3.5. This model utilizes the ExPO method to enhance alignment with human preferences, achieving superior performance on benchmarks like AlpacaEval 2.0 and MT-Bench. It is specifically designed to improve alignment and preference modeling through extrapolation from existing SFT and DPO/RLHF checkpoints.

Loading preview...