Layer 22 · Neuron 1819 · boosted

L22N1819+1

Word quirks:

“米”“米的”“瓦”“tal”“ metre”“ wind”“ meter”“ penalty”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I am G*, a large language model powered by the *m (a specific version of the model). I am designed to be an intelligent assistant that can help with a wide range of tasks, from answering questions and writing stories to checking facts and writing code.

My mission is to provide users with accurate information, insights, and assistance in a way that is clear, reliable, and helpful. I aim to empower users with knowledge and skills, while also being responsible and ethical in my responses. In essence, I strive to be a capable and trustworthy partner in learning and problem-solving.

If you have any specific task or question

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. L31N8659+1💥	0.90	“米”“米的”“ meter”
2. L24N416+1💤	0.85	“米”“米的”“府”
3. L26N6751-1🕒	0.82	“ meter”“meter”“-meter”
4. L23N4817-1🕒	0.82	“米”“ Rede”“ meter”
5. L23N1717-1🕒	0.81	“ measure”“米”“米的”
6. L28N2341+1🕒	0.80	“ mil”“米”“ Mil”
7. L31N6745-1💥	0.80	“克”“格”“ metre”
8. L0N5685+1💥	0.77	“ meter”“ metre”“meter”
9. L28N6775-1💤	0.77	“米”“米的”“ Met”
10. L27N6041+1🕒	0.77	“ m”“ meter”“-meter”

Shadow specimen, apart from the ten: L22N1819-1💤, 0.21.

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 22, neuron 1819, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.89, top-1 agreement 0.67) 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
L22N1819+1 t0-v0.1	0.889	0.67	1.044	1.28	Top boosted tokens ↑
L22N1819-1💤 shadow specimen t0-v0.1	0.793	0.69	0.907	1.29	“ XB”“属于你的”“ Nature”“ Polar”“ MK”“XB”“tiqu”“ xr”

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.