mirau-agent-base-oai
Introduction
mirau-agent-base is a large language model specifically optimized for Agent scenarios, fine-tuned based on Qwen2.5-14B-Instruct. This model focuses on enhancing multi-turn tool calling capabilities, enabling autonomous planning, task execution, and exception handling in complex interactive environments.
Although named "base", this does not refer to a foundation model that has only undergone pre-training, but rather a "cold-start" version that has been through supervised fine-tuning (SFT) and direct preference optimization (DPO), providing a high-quality policy initialization foundation for subsequent reinforcement learning training. We also hope the community can further enhance it through RL training based on this foundation.
The oai version is compatible with OpenAI's function definitions, with no other differences from the previous mirau-agent-14b-base.