AgentTrove: 1.7M-Sample Agentic Dataset Released from OpenThoughts
OpenThoughts has released AgentTrove on HuggingFace, a dataset containing 1.7 million samples designed for training and evaluating AI agents. The dataset provides a large-scale resource for agentic capability development across diverse task types, contributing to the growing open-source ecosystem for agent training data.
Why It Matters
At 1.7 million samples, AgentTrove is one of the larger publicly available agentic training datasets, enabling teams without proprietary data collection infrastructure to train and fine-tune agent-capable models at meaningful scale.