Layer 4 · Neuron 2392 · suppressed

L4N2392-1

Word quirks:

“ matrices”“ matrix”“矩阵”“ Matrix”“ tensor”“ tensors”“matrix”“ vectors”

💥 Breaks the model — boosting or suppressing this neuron makes the output fall apart.

Identity (according to the specimen itself)

This question addresses two fundamental concepts in linear algebra: the identity matrix (denoted as $I$) and the Kronecker product (denoted as $\otimes$).

Identity Matrix ($I$): The identity matrix of a square matrix $A$, denoted as $A_I$, is simply $A$ itself. It is a matrix where the element at position $(i, j)$ is $A_{ij}$.
Kronecker Product ($\otimes$): The Kronecker product of two matrices $A$ and $B$,

Nearest neighbors

Specimens with similar replies or word quirks. Similarity runs from −1 to 1; 1 is identical. These are leads, not shared mechanisms. How profiles are built.

By word quirks Profile recipe

Specimen	Similarity	Word quirks
1. L25N1339+1🕒	0.93	“ matrix”“矩阵”“ matrices”
2. L2N1162+1💥	0.92	“ matrix”“ matrices”“ transpose”
3. L5N1158+1🕒	0.90	“ matrix”“ matrices”“ Matrix”
4. L6N3405+1🕒	0.90	“ matrix”“ matrices”“ Matrix”
5. L2N8145+1💥	0.89	“ matrix”“<\|im_end\|>”“ matrices”
6. L25N5191+1🕒	0.89	“ matrix”“矩阵”“ Matrix”
7. L23N1409+1🕒	0.89	“ matrix”“矩阵”“ Matrix”
8. L7N342+1🕒	0.89	“ matrix”“ matrices”“矩阵”
9. L25N2852-1🕒	0.89	“ Matrix”“ matrix”“Matrix”
10. L8N3348-1💤	0.88	“ matrix”“ matrices”“ Matrix”

Shadow specimen, apart from the ten: L4N2392+1🕒, 0.20.

Cousins · specimens sharing a neuron

No specimen with replies shares this neuron yet.

Technical details

Quick scan: measure prediction shifts on fixed reference text. Short test: generate replies to a small set of prompts. Longer test: extend to a larger set. Full test: run the full prompt suite. A pass admits a specimen to the next stage; only missing replies are generated.

Layer 4, neuron 2392, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 3.04, top-1 agreement 0.36) is outside the band that feeds the short test (KL 1 to 2.5, agreement 0.5 to 0.65), so it stays at the quick scan unless someone queues it by hand. A hand pick goes straight to the full test. See Method for the tiers.

Quick scan · prediction measurements

Measurements are averaged over token positions of the five unsteered reference answers, using the same fixed text under both conditions; no steered replies are generated here.

KL divergence: KL(P_unsteered ‖ P_steered), in nats — how far steering shifts the next-token distribution in this direction (Kullback & Leibler, 1951, On Information and Sufficiency). Unsteered ≈ 0.001.
Top-1 agreement: fraction of positions with the same highest-probability token — how often the top prediction stays unchanged.
ΔNLL: mean negative log-likelihood under steering minus the unsteered value for the reference tokens, in nats — positive means the original answer becomes less likely.
Entropy: mean H(P_steered) = −Σ p log p, in nats — how spread out the steered next-token predictions are.

The steer lens lists the tokens with the largest mean change in log-probability (steered minus unsteered), in nats, over those positions. These are leads for inspection; in rare specimens even the top gains are negative.

Setting	KL divergence	Top-1 agreement	ΔNLL	Entropy	Steer lens
L4N2392+1🕒 shadow specimen t0-v0.1	2.353	0.45	2.767	2.36	“ cosmic”“ nucle”“ stellar”“ cosm”“ Cosmic”“ stars”“ nuclear”“ astro”
L4N2392-1 t0-v0.1	3.039	0.36	3.490	2.28	Top boosted tokens ↑

Method

A specimen is a fixed steering configuration of Qwen3.5-4B: one MLP neuron or a pair, each with a signed dose. Unsteered is the control. L identifies the layer; N, the neuron.

Additive activation steering. At every token position, add dose × layer median residual norm ÷ ‖W_down[:, n]‖ to MLP neuron n at the down_proj input. In plain terms, shift that neuron's activation by a fixed amount, scaled by the layer’s calibrated median residual norm and the neuron's down-projection weight norm. Positive doses push up; negative doses push down. Dose is normalized intervention size, not a percentage or probability.