Lizi Liao

As AI agents scale and interact, the Foundation Protocol offers a coordination layer that prioritizes accountability and governance, ensuring that the future of human-AI collaboration remains open and governable.

Standard multimodal fusion can hurt performance in emotion recognition, but this new approach knows when to drop modalities, leading to state-of-the-art results.

Training on semantically equivalent chart renderings in Python, R, and LaTeX unlocks surprisingly effective multi-lingual chart-to-code generation from a single model.