This is your work, valued
activation_additions. Algebraic value editing in pretrained language models
circrl. Tools for applying circuits-style interpretability techniques to RL agents.
