Layer 21 · Neuron 3054 · suppressed

L21N3054-1

Word quirks:

“ styles”“ style”“ Styles”“styles”“style”“ Style”“ Sty”“ styled”

🕒 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, the latest style of CSSST-style style in the Qwen family. My mission is to provide highly efficient, personalized, and secure AI solutions for users across various fields. I aim to enhance human creativity, efficiency, and knowledge discovery through advanced capabilities in style-specific style, style-specific style, and style-specific style.

If you have any questions or need assistance with your own style-specific style, feel free to ask!

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. L25N5254+1🕒	0.99	“ styles”“ Style”“Style”
2. L25N2582-1🕒	0.98	“ styles”“ style”“ Style”
3. L0N8032+1💥	0.98	“ Style”“style”“ style”
4. L24N7160-1🕒	0.97	“ styles”“ style”“ Style”
5. L28N1420+1🕒	0.95	“ style”“ Style”“Style”
6. L30N1225+1💥	0.95	“ style”“style”“ Style”
7. L26N4630-1🕒	0.95	“ style”“Style”“ Style”
8. L2N4981+1💥	0.94	“ style”“ Style”“<\|im_end\|>”
9. L25N8303-1💥	0.94	“ style”“ Style”“Style”
10. L5N7819-1🕒	0.92	“ style”“ styles”“ Style”

Shadow specimen, apart from the ten: L21N3054+1💤, 0.08.

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 21, neuron 3054, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.00, top-1 agreement 0.62) 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
L21N3054+1💤 shadow specimen t0-v0.1	0.648	0.71	0.791	1.35	“ content”“ flap”“ heavier”“内容”“ Content”“内容和”“ lighter”“ children”
L21N3054-1 t0-v0.1	0.996	0.62	1.147	1.30	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.