Layer 23 · Neuron 1892 · boosted

L23N1892+1

Word quirks:

“文本”“ text”“ Text”“ texts”“textarea”“ textarea”“ TEXT”“TEXT”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am X*, a large text editor (a multi-paragraph text editor) created by the *Hermegene team. My main goal is to help users like you process and generate text in a more efficient way.

Here's how I work:

Text Processing: I can edit, format, and clean up text documents.
Multi-Paragraph Editing: Unlike traditional editors that focus on single lines of text, I can handle multi-paragraph texts, allowing for better editing of long documents.
Efficiency: I aim to provide a faster and more streamlined

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. L31N8180-1💥	0.96	“文本”“ Text”“-text”
2. L26N5196-1🕒	0.95	“text”“ Text”“Text”
3. L25N5068+1🕒	0.95	“文本”“ text”“ texts”
4. L27N4836-1🕒	0.95	“ Text”“文本”“ text”
5. L29N7010+1🕒	0.95	“ text”“文本”“ Text”
6. L25N8272-1🕒	0.95	“ text”“文本”“text”
7. L27N3851+1🕒	0.95	“text”“文本”“ text”
8. L24N4713+1🕒	0.95	“ text”“文本”“.text”
9. L26N2749-1🕒	0.95	“ text”“文本”“ Text”
10. L30N2637-1🕒	0.95	“text”“ text”“文本”

Shadow specimen, apart from the ten: L23N1892-1💤, 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 23, neuron 1892, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L23N1892+1 t0-v0.1	1.393	0.62	1.382	1.29	Top boosted tokens ↑
L23N1892-1💤 shadow specimen t0-v0.1	0.718	0.70	0.821	1.24	“ лу”“ppy”“举”“亮”“而易”“ Burlington”“trer”“컷”

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.