3 papers
cs.LG2026
Interpreting Reinforcement Learning Agents with Susceptibilities
Chris Elliott, Einar Urdshals, David Quarel +1
Susceptibilities are a technique for neural network interpretability that studies the response of posterior expectation values of observables to perturbations of the loss. We gener…
cs.LG2026
Stagewise Reinforcement Learning and the Geometry of the Regret Landscape
Chris Elliott, Einar Urdshals, David Quarel +2
Singular learning theory characterizes Bayesian learning as an evolving tradeoff between accuracy and complexity, with transitions between qualitatively different solutions as samp…
cs.LG2025
SCALAR: Benchmarking SAE Interaction Sparsity in Toy LLMs
Sean P. Fillingham, Andrew Gordon, Peter Lai +3
Mechanistic interpretability aims to decompose neural networks into interpretable features and map their connecting circuits. The standard approach trains sparse autoencoders (SAEs…