Tilde Research studies how AI models work internally and uses those insights to improve model design and training. Its research connects interpretability—the study of a model’s internal mechanisms—with new architectures and pretraining methods.
See something inaccurate or outdated?
