MATS
Mentorship output is intentionally diverse across subfields, which does not collapse into a single unified measurement spine—participants may advance interpretability, control, or governance lines without resolving cross-cutting bridge composition.
Introduction
MATS runs a twelve-week mentored research program pairing scholars with alignment mentors for empirical and conceptual projects across transparency, security, and related subfields. Fellowship output is intentionally diverse and does not collapse into a single unified measurement spine. The program feeds talent into the broader Field hub while individual projects may advance lines such as Inner Alignment without resolving cross-cutting bridge composition.
Who carries it: MATS (Machine Alignment, Transparency, and Security)
What they aim to do. Connect promising scholars to alignment mentors so they can produce substantive empirical or conceptual research during an intensive fellowship term.
The hard question. Mentorship output is intentionally diverse across subfields, which does not collapse into a single unified measurement spine—participants may advance interpretability, control, or governance lines without resolving cross-cutting bridge composition.
What they produce. MATS runs a twelve-week mentored research program with an extension track, pairing scholars with alignment mentors for empirical and conceptual projects.
Key terms. Key terms include MATS, mentored research, mechanistic interpretability, AI control, and related technical subfields across transparency and security.
What they contribute. A research talent pipeline that feeds labs, orgs, and independent projects with mentored early-career alignment work.
How this project treats it. Fellowship research quality is independent of whether typed bridges compose under explicit adversarial-verifiability assumptions.
Links
Map clustering
AISafety.com map listings that roll up to this agenda:
See the coverage matrix for evidence tagged to this agenda, and the glossary for shared terms.