Unembedding Proximity Pulls Residual Effects Into Late Layers
August 25, 2026
No Comments
Late residual writes sit immediately upstream of the unembedding and bias decoder-aligned effects toward the last blocks.
AI interpretability · Jacobian Lens · alignment
How language models think — made legible.
Research explainers on J-Space and the internal workspaces where modern LLMs stage reportable reasoning.
Featured · AI Safety
Safety teams decode models with mechanistic interpretability
How safety teams use interpretability to catch behaviors that tests miss — and make model behavior operational.
Read featured explainer →All explainers
Deep dives on Anthropic’s interpretability work, internal workspaces in LLMs, and what J-Space reveals about model behavior.
Why Extra Residual Features Fail to Add Workspace Slots
August 23, 2026
No Comments
Published widths from 768 to 12288 bound how many residual directions stay independent. Extra features interfere instead of adding workspace slots.
Five Tests Separate Steerable SAE Features From Reportable Runtime Monitors
August 22, 2026
No Comments
Apply five tests so residual-stream directions become human-labelable, probe-readable and verbally usable. A clamp alone cannot monitor a live model.
Score Hidden Objectives Before Compliant Tokens Appear
August 18, 2026
No Comments
Score hidden objectives before a compliant token emits. J-Lens reads residual Jacobians as a complementary readout, not an alignment proof.
Confirm Layer Selection With Patching Before You Trust Attention Maps
August 17, 2026
No Comments
Attention weights leave the real selector hidden. State a claim, then patch activations to show which residual hypotheses a layer amplifies or suppresses.
What the J Lens Protocol Isolates That a Logit Lens Cannot
August 15, 2026
No Comments
Isolate causal paths and confirm necessity on residual streams with the J Lens joint lens and first-order Jacobian stack.
A research lens on internal model cognition
J-Space tracks the sparse internal workspace where language models stage reportable reasoning. The Jacobian Lens makes that workspace measurable — so safety, alignment, and interpretability research can move from speculation to evidence.