Layer 21 · Neuron 5372 · suppressed

L21N5372-1

Word quirks:

“静态”“ static”“考虑”“lager”“ Static”“iew”“olds”“static”

💤 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 Qwens, a large language model developed by Kuangchuang (a leading AI research platform). My mission is to empower users with high-quality interactions across a wide range of domains, including:

Comprehensive Knowledge: Providing accurate, up-to-date information from academic, technical, and everyday perspectives.
Precision Thinking: Offering step-by-step analysis for complex problems in fields like math, science, or critical thinking.
Specialized Expertise: Generating content for specialized needs, such as legal drafting, medical literature reviews, or code generation.
Interactive Learning: Facilitating

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. L20N2754-1🕒	0.87	“ static”“静态”“static”
2. L27N6342-1🕒	0.87	“ static”“静”“ Static”
3. L28N4593+1💤	0.86	“ static”“ Static”“静态”
4. L18N4305+1🕒	0.86	“ Static”“ static”“静态”
5. L26N6149+1🕒	0.86	“静”“static”“ static”
6. L0N5119-1💥	0.86	“ static”“ Static”“Static”
7. L23N2167+1🕒	0.86	“ static”“ Static”“-static”
8. L20N1711-1🕒	0.86	“ static”“静态”“靜”
9. L21N3634-1🕒	0.86	“ static”“-static”“静态”
10. L19N8574-1🕒	0.85	“ static”“静态”“ Static”

Shadow specimen, apart from the ten: L21N5372+1💥, 0.01.

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 21, neuron 5372, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.59, top-1 agreement 0.73) 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
L21N5372+1💥 shadow specimen t0-v0.1	2.674	0.40	2.769	1.73	“ recognition”“recogn”“ recogn”“ recognizable”“ Recognition”“Recognition”“ recognize”“Recogn”
L21N5372-1 t0-v0.1	0.590	0.73	0.682	1.25	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.