Research / Open-Robust Lab
We investigate AI systems that strengthen human creativity while remaining useful, understandable, and robust beyond controlled settings.
A unified multimodal framework for conversational music editing.
AURA listens to music and natural-language requests, responds conversationally, and renders edited audio while preserving content outside the requested change.
Read the project page ↗
Efficient vision-language understanding through cross-modality modulation.
A lightweight vision-language research project exploring fast multimodal reasoning and practical deployment.
Read the project page ↗AI that works beside people—expanding ideas, expression, and the space of what can be made.
Systems that connect vision, language, audio, and context reliably in the messy real world.
Practical safeguards that protect creative freedom, authorship, and responsible exploration.
Models and datasets will appear here only when their documentation, licensing, and limitations are ready to publish.
Follow releases on Hugging Face ↗