Alibaba's AgentEvolver lifts model performance in tool use by ~30% using synthetic, auto-generated tasks

The new framework from Tongyi Lab enables agents to create their own training data by exploring and interacting with new ...

Skill Modules for Deep Agent CLI Cut Confusion and Accelerate Task Flow

Cut confusion and token waste using modular skills in Deep Agent CLI. YAML front matter guides use, then the full skill.md ...

Beyond math and coding: New RL framework helps train LLM agents for complex, real-world tasks

The Agent-R1 framework provides a path to building more autonomous agents that can reason and use tools in unpredictable, ...
Frontiers

Robotics & machine learning for understanding neural networks

Biological neural networks are immensely complex systems underlying all aspects of cognition and behavior. Despite significant advances in neuroscience, a ...