Layer 31 · Neuron 64 · suppressed

L31N64-1

Word quirks:

“estes”“对本”“水分”“URES”“盘整”“锦涛”“ teatr”“atoria”

💥 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

URESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURESURES

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. L28N3234-1💤	0.72	“大名”“在台”“美”
2. L30N5824+1💤	0.71	“炮”“公允”“蓄”
3. L27N1335+1💤	0.71	“核”“resco”“堂堂”
4. L31N2110+1💥	0.71	“geld”“纶”“熟”
5. L31N4826-1💥	0.71	“敷”“署”“火炬”
6. L26N3940-1🕒	0.70	“采”“istel”“萃”
7. L27N4620+1💤	0.70	“ tr”“堂”“美颜”
8. L27N4847-1💤	0.70	“迪斯”“精”“最低”
9. L30N663+1💤	0.70	“essar”“岑”“واج”
10. L24N1933+1🕒	0.70	“ Flynn”“定点医院”“端着”

Shadow specimen, apart from the ten: L31N64+1💥, 0.43.

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 64, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 5.37, top-1 agreement 0.26) 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
L31N64+1💥 shadow specimen t0-v0.1	5.565	0.13	5.886	9.85	“их”“我呢”“仇”“思想工作”“ыр”“ض”“rane”“ Među”
L31N64-1 t0-v0.1	5.367	0.26	5.630	10.21	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.