Develops open-weight Cogito language models that can answer directly or reason before responding. Its research explores iterative self-improvement—training models to internalize useful reasoning rather than relying only on longer inference-time search.
See something inaccurate or outdated?
