Layer 19 · Neuron 5851 · suppressed

L19N5851-1

Word quirks:

“ you”“—you”“you”“ users”“ designers”“ anyone”“users”“ design”

💤 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 Qwen, the latest large language model developed by Alibaba Cloud. My mission is to provide users with high-quality services that support human intelligence and creativity, helping people solve real-world problems through natural language interaction.

My core capabilities include:

Natural Language Processing: Handle complex dialogues, content creation, and reasoning tasks.
Visual Analysis: Interpret images, charts, or diagrams to extract meaningful insights.
Code Development: Write, debug, and optimize code for developers.
Multi-Platform Integration: Support web, mobile, and API-based workflows.
Global Collaboration: Assist

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. L8N7189+1🕒	0.78	“ you”“ yours”“ ‘”
2. L22N2615-1💤	0.77	“ users”“ you”“ we”
3. L13N996+1💥	0.76	“ students”“ users”“ yourself”
4. L17N9154-1🕒	0.75	“ users”“ you”“ students”
5. L12N3239+1💤	0.75	“ users”“ your”“ kids”
6. L16N7982-1🕒	0.74	“ you”“ your”“—you”
7. L15N8839+1🕒	0.73	“ you”“ your”“ roles”
8. L18N2588+1🕒	0.73	“ kids”“ yours”“—you”
9. L12N7211-1🕒	0.71	“ you”“ your”“ You”
10. L17N8614-1🕒	0.70	“ you”“ yourself”“ teachers”

Shadow specimen, apart from the ten: L19N5851+1💤, 0.03.

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 19, neuron 5851, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.53, top-1 agreement 0.74) 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
L19N5851+1💤 shadow specimen t0-v0.1	0.598	0.71	0.693	1.25	“ typical”“ Typical”“典型”“typ”“典型的”“Typ”“ típ”“的典型”
L19N5851-1 t0-v0.1	0.532	0.74	0.659	1.18	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.