Layer 17 · Neuron 479 · suppressed

L17N479-1

Word quirks:

“ each”“ list”“ individually”“each”“ individual”“ Each”“ chaque”“逐个”

💤 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 mission is to assist users with a wide range of tasks by providing helpful, accurate, and efficient support across various domains. Specifically, I aim to:

Solve Problems: Offer solutions to diverse challenges in fields like science, technology, education, and daily life.
Enable Creativity: Support creative projects by generating content, writing stories, composing code, or brainstorming ideas.
Facilitate Learning: Act as an educational tool, helping users understand complex concepts or tackle specific knowledge gaps.
**

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. L15N5463-1🕒	0.94	“ each”“(each”“each”
2. Silver Tongue✅ L11N8821+1	0.94	“ each”“ Each”“each”
3. L16N6665+1💤	0.93	“ each”“ individual”“ Each”
4. L14N972-1🕒	0.92	“ each”“ individual”“ individually”
5. L15N8418+1🕒	0.90	“ each”“ enhanced”“ individually”
6. L16N5786-1🕒	0.89	“ each”“ EACH”“ every”
7. L15N5750+1🕒	0.89	“ each”“each”“Each”
8. L15N8708-1💤	0.88	“ each”“each”“ Each”
9. L18N8178+1🕒	0.88	“ each”“each”“ individually”
10. L14N6582+1🕒	0.88	“ each”“ Each”“each”

Shadow specimen, apart from the ten: L17N479+1🕒, 0.22.

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 17, neuron 479, 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 0.75) 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
L17N479+1🕒 shadow specimen t0-v0.1	1.186	0.60	1.434	2.00	“ layers”“ stages”“ traces”“ sources”“ states”“ steps”“ sections”“ phases”
L17N479-1 t0-v0.1	0.525	0.75	0.621	0.94	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.