Add nonlinear probes, streaming extraction, and mediation analysis#8

Merged

lgngrvs merged 2 commits into2.1from

worktree-interpretability-work

Apr 10, 2026

Commits on Apr 3, 2026

Add nonlinear probes, class reweighting, streaming extraction, and mediation analysis

lgngrvs
and
claude
committed

Commits on Apr 4, 2026

Add subspace activation patching experiment via interchange interventions

lgngrvs
and
claude
committed