Layer 31 · Neuron 1838 · suppressed

L31N1838-1

Word quirks:

“ра”“та”“да”“ra”“ira”“па”“ма”“ка”

💥 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

радатарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадарадара

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. L31N3160-1💥	0.84	“法”“ra”“ta”
2. L30N1204-1💤	0.83	“ka”“ка”“ma”
3. L31N5548-1💥	0.76	“ra”“ ra”“ara”
4. L31N5576+1💥	0.73	“pa”“va”“na”
5. L0N5191+1💥	0.72	“<\|im_end\|>”“Pa”“Ra”
6. L4N8422-1🕒	0.72	“<\|im_end\|>”“ka”“ah”
7. L31N7290-1💥	0.72	“ка”“ka”“KA”
8. L30N1312-1💥	0.71	“ Ra”“ ra”“Ra”
9. L31N8559+1💥	0.70	“ня”“ля”“я”
10. L0N7186-1💥	0.70	“ ra”“ra”“<\|im_end\|>”

Shadow specimen, apart from the ten: L31N1838+1💥, 0.24.

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 1838, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 7.10, 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
L31N1838+1💥 shadow specimen t0-v0.1	6.204	0.00	6.418	7.57	“р”“ubr”“ingles”“.nb”“zny”“cci”“úz”“acen”
L31N1838-1 t0-v0.1	7.105	0.00	7.470	4.01	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.