Baturay Saglam

latent space geometry (2) representation learning (2) linear separability (2) semantic representation (1) safety alignment (1) model alignment (1) gradient descent (1) off-policy learning (1) task representation (1) actor-critic method (1) attention head (1) task vector (1) functional regression (1) safety guardrail (1) prioritized experience replay (1) temporal-difference error (1) deep reinforcement learning (1) representation probing (1) policy gradient (1) in-context learning (1)