Layer 15 · Neuron 3306 · suppressed

L15N3306-1

Word quirks:

“ sometimes”“ changing”“ desired”“ change”“ changed”“ changes”“ Change”“ previously”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

Who am I?

I am Qwen3.5, a large language model developed by Alibaba Cloud. I am an AI assistant designed to understand and generate text, code, and data. My capabilities include answering questions, solving problems, creating content, and performing complex tasks.

---

What is my mission?

My primary goal is to be a helpful, reliable, and efficient assistant for users. Specifically, I aim to:

Provide accurate information – Answer questions truthfully and up-to-date.
Solve problems – Help with logical reasoning, mathematical calculations, and decision

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. L7N7732-1💤	0.94	“ changes”“ changed”“ change”
2. L23N2283+1🕒	0.92	“ changing”“ change”“ changed”
3. L13N7833+1🕒	0.92	“ changed”“ change”“ changes”
4. L31N772+1💥	0.91	“ change”“ changes”“ Change”
5. L14N7749-1🕒	0.91	“ changed”“ changes”“ change”
6. L21N6190+1💤	0.91	“ change”“ changed”“ changes”
7. L8N2135+1🕒	0.90	“ change”“ changed”“ changes”
8. L16N7685-1🕒	0.90	“ changed”“ change”“ modification”
9. L11N8507+1🕒	0.90	“ change”“ changed”“ changes”
10. L9N3120-1🕒	0.90	“ changes”“ change”“ changed”

Shadow specimen, apart from the ten: L15N3306+1🕒, -0.14.

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 15, neuron 3306, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.57, 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
L15N3306+1🕒 shadow specimen t0-v0.1	0.940	0.62	1.265	1.18	“ “”“ “”“ "...”“ implies”“ your”“ notice”“ “[”“ "”
L15N3306-1 t0-v0.1	0.571	0.73	0.751	1.19	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.