Layer 31 · Neuron 4112 · boosted

L31N4112+1

Word quirks:

“hoz”“hle”“strup”“ult”“quem”“ity”“ettes”“adurch”

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

Identity (according to the specimen itself)

The reply didn't hold together

ityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityityity

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. L29N1003-1💤	0.69	“udiant”“ HG”“uiten”
2. L29N5536+1💤	0.68	“ele”“idu”“adesh”
3. L28N248+1💤	0.68	“essler”“ Naz”“hra”
4. L30N5912-1💤	0.67	“icht”“íben”“kep”
5. L19N1402+1💤	0.67	“ce”“kest”“ange”
6. L31N8662-1💥	0.67	“hov”“et”“uth”
7. L29N996-1💤	0.67	“ze”“els”“pero”
8. L31N8720-1💥	0.67	“bies”“ch”“vern”
9. L30N6000+1💤	0.66	“bek”“und”“ulu”
10. L27N590-1💤	0.66	“esel”“ibilitas”“uchs”

Shadow specimen, apart from the ten: L31N4112-1💥, 0.38.

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 31, neuron 4112, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 6.74, top-1 agreement 0.00) 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
L31N4112+1 t0-v0.1	6.745	0.00	7.074	6.34	Top boosted tokens ↑
L31N4112-1💥 shadow specimen t0-v0.1	9.051	0.00	9.303	2.31	“ypes”“есь”“ting”“arget”“ech”“ype”“YPE”“ติ้ง”

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.