Token-time interpretability instrument for language models — measures how generation trajectories move, branch, and respond to perturbation. Deterministic intervention comparisons via SeedCache branchpoints, hysteresis protocol, and SAE-feature steering.
transformers pytorch language-models ai-safety sae interpretability mechanistic-interpretability nnsight activation-steering perturbation-experiments
-
Updated
Apr 30, 2026 - Python