Layer 31 · Neuron 8200 · suppressed

L31N8200-1

Word quirks:

“計”“梢”“都好”“着小”“络绎”“有此”“麒”“ева”

💥 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

I am **IARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARARAR

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. L31N475-1💥	0.88	“屿”“另”“些”
2. L23N4780-1💤	0.87	“井”“所向”“在所”
3. L31N2103+1💥	0.87	“墒”“京”“两口”
4. L30N2391+1🕒	0.87	“委”“胜”“尔克”
5. L27N6606-1💤	0.87	“靠”“草”“我去”
6. L23N31-1💥	0.87	“自”“合”“欧”
7. L28N898+1🕒	0.87	“不起”“签”“琦”
8. L31N2663-1💥	0.87	“陀”“谷”“吉”
9. L31N2579-1💥	0.86	“恙”“疙”“璀”
10. L27N6485-1🕒	0.86	“溢”“辛”“爆”

Shadow specimen, apart from the ten: L31N8200+1💥, 0.06.

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 8200, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 4.93, top-1 agreement 0.29) 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
L31N8200+1💥 shadow specimen t0-v0.1	8.287	0.00	8.514	4.76	“uli”“azz”“ндекс”“acobian”“ordan”“ریان”“ensen”“upiter”
L31N8200-1 t0-v0.1	4.930	0.29	5.285	10.03	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.