Layer 20 · Neuron 9126 · suppressed

L20N9126-1

Word quirks:

“ anything”“ any”“ everything”“anything”“的任何”

🕒 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 (also known as Tongzi), a large language model developed by Alibaba Cloud. My mission is to provide any knowledge or information that you need, including but not limited to:

Knowledge and Insights: Any questions you ask, I will answer with the most accurate and relevant information.
Learning and Exploration: Any learning or exploration you do, I will help you find the results and insights you get.
Communication and Interaction: Any communication or interaction you have, I will ensure that everything you say is clear and understandable.
**Creative and Practical

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. L22N8557-1🕒	0.96	“任何”“ anything”“ any”“any”“anything”
2. L18N3371+1🕒	0.95	“ any”“ anything”“任何”“ ANY”“any”
3. L16N7967+1💤	0.93	“ any”“ anything”“ ANY”“ every”“任何其他”
4. L16N8060-1🕒	0.93	“ any”“任何”“ anything”“ anywhere”“ ANY”
5. L16N8794+1🕒	0.93	“ anything”“ any”“ except”“ Any”“ ANY”
6. L21N7569-1🕒	0.93	“ anything”“ Anything”“anything”“Anything”“ anybody”
7. L18N3119+1💤	0.93	“ any”“任何”“ Any”“的任何”“ anywhere”
8. L20N4066+1🕒	0.92	“ anything”“ any”“anything”“有任何”“任何东西”
9. One-Thing Attractor✅ L18N6903+1	0.92	“one”“thing”“learned”“teach”“lesson”
10. L18N2092-1🕒	0.91	“ anywhere”“ anything”“ any”“任何形式的”“ whatsoever”

Shadow specimen, apart from the ten: L20N9126+1💤, 0.20.

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 9126, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.95, 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
L20N9126+1💤 shadow specimen t0-v0.1	0.405	0.74	0.445	1.11	“ possibility”“ possibilities”“月中”“的可能”“的可能性”
L20N9126-1 t0-v0.1	0.945	0.62	1.220	1.38	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.