Work 01
LegoFlow
A fully agentic data workflow connecting curation, trajectory generation, model training, and evaluation in one extensible loop.
An open-source collection to advance frontier agentic intelligence
LegoX is a series of open-source projects spanning agentic data pipelines, training frameworks, evaluation benchmarks and foundation models. Our mission is to connect these pieces into reproducible and extensible systems that advance frontier intelligence.
2026-08-01
We introduce LegoFlow, a fully agentic data workflow that connects data curation, trajectory generation, model training, and evaluation in one extensible loop.
Work 01
A fully agentic data workflow connecting curation, trajectory generation, model training, and evaluation in one extensible loop.
Work 02
A supervised fine-tuning recipe and model family for resolving real software issues with carefully validated trajectories.
Work 03
An agentic code-review system that diagnoses repository-level problems and closes the generate–review–revise loop.
Work 04
A study of interaction trajectories and environment-grounded supervision for building more capable terminal agents.
Collaborating institutions