Layer 20 · Neuron 5583 · suppressed

L20N5583-1

Word quirks:

“ spreads”“ spreading”“ spread”“ width”“ narrower”“ wider”“Spread”“ Spread”

🕒 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 Qwen3.3, the latest version of the QianZhou (TianXin) series developed by Microsoft. My mission is to help users achieve their goals by providing high-quality interactions across a wide range of domains. Here's how I approach this:

Core Mission:

Broad Knowledge & Precision: Leverage my advanced knowledge base to deliver accurate, context-aware responses in fields like science, technology, arts, and daily life.
Adaptive Learning: Interact with users to understand their needs, whether they're asking for quick answers, complex problem-solving,

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. L17N8393+1🕒	0.94	“ spreading”“ spread”“ spreads”
2. L25N3495-1🕒	0.91	“ spread”“ Spread”“Spread”
3. L27N6698+1🕒	0.89	“ spread”“ Spread”“Spread”
4. L30N1953+1💥	0.89	“ spread”“ Spread”“spread”
5. L26N3380-1🕒	0.89	“ spreading”“ spread”“ spreads”
6. L22N5955+1🕒	0.88	“ spread”“ spreading”“ spreads”
7. L25N9089+1💤	0.88	“ spread”“ spreads”“ Spread”
8. L2N5834-1💥	0.87	“ spreading”“ spread”“<\|im_end\|>”
9. L21N3638-1💤	0.87	“ spread”“ Spread”“ stretch”
10. L29N5245+1🕒	0.87	“ spread”“spread”“ spreads”

Shadow specimen, apart from the ten: L20N5583+1💤, 0.10.

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 20, neuron 5583, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.68, top-1 agreement 0.68) 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
L20N5583+1💤 shadow specimen t0-v0.1	0.621	0.70	0.747	1.19	“ dét”“毒蛇”“ detecting”“ Detect”“ fraudulent”“cord”“ silent”“加速”
L20N5583-1 t0-v0.1	0.682	0.68	0.770	1.18	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.