UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3
TEXT GENERATIONConcurrency Cost:1Model Size:9BQuant:FP8Ctx Length:16kPublished:Jun 29, 2024License:gemmaArchitecture:Transformer0.1K Warm

UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3 is a 9 billion parameter Gemma-2-9B-It-based causal language model developed by UCLA-AGI. This model is the third iteration fine-tuned using Self-Play Preference Optimization (SPPO) on synthetic datasets, primarily in English. It is optimized for improved alignment and response quality, demonstrating enhanced win rates on the AlpacaEval Leaderboard compared to previous iterations.

Loading preview...