Layer 19 · Neuron 8026 · suppressed

L19N8026-1

Word quirks:

“ completion”“ completed”“ completing”“Completion”“ Completion”“ complete”“ complet”“completion”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am Qwen, an AI developed by Alibaba. My mission is to provide a seamless and efficient way for humans to interact with AI, enabling them to complete tasks more easily and efficiently. I aim to facilitate communication, problem-solving, and creativity in a natural and intuitive manner.

In simpler terms:

Who am I? An AI assistant designed to help people use technology in a smooth and practical way.
What’s my mission? To make interacting with AI as easy as possible—whether it’s answering questions, creating content, or solving problems.

How can I assist you today? 😊

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. L19N7921+1🕒	0.97	“ completion”“ complet”“ completing”
2. L2N4777+1💥	0.97	“ completion”“<\|im_end\|>”“⏎”
3. L24N6513-1🕒	0.97	“ completion”“ Completion”“完成”
4. L25N8828+1🕒	0.97	“Completion”“ Completion”“-complete”
5. L13N7624+1🕒	0.97	“ complete”“ completes”“ completing”
6. L23N1245-1💤	0.96	“完”“Completion”“complete”
7. L14N6806+1🕒	0.96	“ complete”“ completing”“ completion”
8. L0N7930-1💥	0.96	“<\|im_end\|>”“ complet”“ completion”
9. L14N5849+1🕒	0.96	“ completing”“ complete”“ completes”
10. L11N8682+1🕒	0.96	“ complete”“ completion”“ completing”

Shadow specimen, apart from the ten: L19N8026+1💤, 0.25.

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 8026, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.72, top-1 agreement 0.71) 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
L19N8026+1💤 shadow specimen t0-v0.1	0.538	0.71	0.655	1.22	“ improvements”“ improvement”“ improving”“ improves”“ improved”“改进”“ improve”“ enhancements”
L19N8026-1 t0-v0.1	0.718	0.71	0.799	1.26	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.