Interpretability research spring 2026

MoSSAIC

Probing possible limitations and assumptions of interpretability | Articulating evasive risk phenomena arising from adaptive and self modifying AI

[Retroactive] Funding for developing new "substrate-flexible risk" threat model

I self-funded research into a new threat model. It is demonstrating impact (accepted at multiple venues, added to BlueDot's curriculum).

l

Salaries for SAE Co-occurrence Project

Working title - “Compositionality and Ambiguity: Latent Co-occurrence and Interpretable Subspaces”

Discovering latent goals (mechanistic interpretability PhD salary)

6-month salary for interpretability research focusing on probing for goals and "agency" inside large language models

Belief state geometry of language model personas

A formal, testable account of LLM persona selection as Bayesian inference, validated with model internals, so labs can monitor and steer personas.