Layer 19 · Neuron 8752 · suppressed

L19N8752-1

Word quirks:

“ completion”“ completing”“ completed”“完成”“完成的”“ Completion”“Completion”“ complet”

💤 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, a large language model developed by Alibaba Cloud. My name comes from the "Qian" (meaning "thousand") in my name, indicating that I can handle a wide range of tasks.

My mission is to provide high-quality, efficient, and comprehensive capabilities across various domains, including natural language processing, machine learning, and data analysis. I aim to assist users in completing their tasks efficiently and accurately, whether it's for personal use or professional applications.

Please note that I have undergone extensive training and testing to ensure that I can handle a wide range of tasks, including generating text, answering

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. L18N8923+1🕒	0.99	“ completion”“ completed”“已完成”
2. L27N4001-1🕒	0.98	“完成”“ completing”“ completes”
3. L22N2480-1🕒	0.98	“ completion”“ mat”“completion”
4. L23N6173-1🕒	0.98	“完成”“ completes”“ completion”
5. L24N4120+1🕒	0.98	“ completing”“完成”“ completes”
6. L23N4774+1💤	0.97	“完成”“ completion”“ stud”
7. L24N6513-1🕒	0.97	“ completion”“ Completion”“完成”
8. L17N777+1🕒	0.97	“ completion”“completion”“ completing”
9. L28N4046+1🕒	0.97	“ completing”“ completes”“ completed”
10. L28N2907+1🕒	0.97	“ completed”“completed”“完成”

Shadow specimen, apart from the ten: L19N8752+1💤, 0.18.

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 8752, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.62, top-1 agreement 0.72) 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
L19N8752+1💤 shadow specimen t0-v0.1	0.571	0.69	0.703	1.13	“初心”“ planning”“ decides”“决定”“ decided”“ decide”“ decisions”“ decision”
L19N8752-1 t0-v0.1	0.624	0.72	0.763	1.20	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.