Layer 18 · Neuron 8002 · suppressed

L18N8002-1

Word quirks:

“ twice”“ two”“ TWO”“ Twice”“ doubled”“ duplicated”“ multiple”“two”

🕒 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, a large-scale language model trained by Alibaba Cloud. I have been trained multiple times with high-quality data in two different knowledge periods, which means I can answer questions accurately from both past and present. My training data is very rich, and I have good performance in multiple languages, including Chinese, English, and even more than 200 languages.

My mission is to help you solve problems through my capabilities in multiple fields. I can assist you in various ways:

Multiple Knowledge Scenarios: I can answer questions about knowledge, code, or even creative writing.
**Multiple

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. L21N8112-1🕒	0.94	“ twice”“ doubled”“ TWO”
2. L19N3941+1💤	0.90	“ two”“ twice”“ TWO”
3. L22N7217-1💤	0.87	“ twice”“ Twice”“ double”
4. L19N6555+1💤	0.86	“ TWO”“ doubles”“ double”
5. L14N5269+1💤	0.86	“ two”“ twin”“ TWO”
6. L14N8508-1🕒	0.85	“ two”“ TWO”“two”
7. L24N9084+1🕒	0.85	“两倍”“ two”“ TWO”
8. L11N7788-1🕒	0.85	“ twice”“ two”“ dual”
9. L19N4975+1🕒	0.85	“ two”“ doubles”“two”
10. L21N6009-1🕒	0.84	“ two”“ Two”“Two”

Shadow specimen, apart from the ten: L18N8002+1💤, 0.03.

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 18, neuron 8002, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.82, top-1 agreement 0.64) 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
L18N8002+1💤 shadow specimen t0-v0.1	0.497	0.72	0.619	1.16	“ exact”“立刻”“ instantly”“新朋友”“hofer”“ Exact”“ partner”“ุง”
L18N8002-1 t0-v0.1	0.823	0.64	0.907	1.23	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.