aiwaves-cn/agents

Project overview

It brings neural-network-inspired training techniques—backpropagation, loss functions, and weight updates—into the symbolic domain of language agents, directly addressing the historical lack of a systematic framework for agent optimization.

Project type

AI Agent · Model Development

Use cases

Automation

Deployment

Refer to project documentation

License

Apache-2.0

Best for

• Researchers and AI engineers looking to systematically train and optimize the symbolic components of language agents.

Key capabilities

• A systematic framework for training language agents using language-based loss, gradients, and weights to optimize prompts and tools.
• Back-propagates language loss along the agent trajectory to generate textual analyses and reflections for symbolic components.
• Updates symbolic components and the computational graph based on language gradients.
• Supports optimizing multi-agent systems by considering nodes as different agents or allowing multiple agents to take actions in one node.

Getting started

• Install the library directly via pip using the command: pip install git+https://github.com/aiwaves-cn/agents@master.

Evidence and sources

• GitHub project description: An Open-source Framework for Data-centric, Self-evolving Autonomous Language Agents
• README: 
• README: Agent symbolic learning is a systematic framework for training language agents, which is inspired by the connectionist learning procedure used for training neural nets. We make an…
• README: pip install git+https://github.com/aiwaves-cn/agents@master
• README: If you find our repository useful in your research, please kindly consider cite:

Reviewed at: 2026-07-16T13:46:40Z