kapynResearch

Environment-free Synthetic Data Generation for API-Calling Agents

Environment-free synthetic data generation removes the need for executable APIs when training LLM agents. The method uses large language models as on-the-fly digital world models to generate realistic interaction trajectories from basic API specifications alone. This approach solves a major scaling bottleneck for developers building tool-using and API-calling AI agents.

Apple ML Research·Jul 21, 2026

Opening Kapyn…