v-stone/affine-vic-9

This model is a novel method to evolve generall-capable LLM-based agents across multiple environments. This model first trains a base generally-capable agent with behavioral cloning, equipping it with basic abability and prior knowledgs. Subsequently, the agent is allowed to perform exploration and learning acorss various tasks and environments.

This model is trained with the This model algorithm on Llama-2-Chat-7B. The model is first trained on the AgentTraj set with behavioural cloning. Next it performs exploration and learning from a broader set of instructions. After evolution, its performance outperforms SOTA models on many tasks.