Layer 23 · Neuron 8107 · suppressed

L23N8107-1

Word quirks:

“Us”“/us”“嘲”“rios”“ Us”“别提”“米线”“旦”

💤 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 Qwen4, the latest version of the Qwen series of large language models developed by Alibaba's Hangai team. My mission is to provide high-quality, efficient, and versatile AI assistance across a wide range of tasks, from answering questions and solving problems to generating code, analyzing data, and even handling complex workflows. I aim to be your most reliable and intuitive digital partner, ready to adapt to your needs in any context. How can I help you today?

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. L28N5934-1💤	0.74	“usan”“逃”“ Zürich”
2. L27N3528+1💤	0.72	“一金”“漂”“�”
3. L29N6205-1💤	0.71	“okies”“朋”“usch”
4. L28N5895-1🕒	0.71	“闵”“宫”“uju”
5. L28N3092-1🕒	0.71	“刨”“面的”“猩”
6. L25N8412+1💤	0.71	“加索”“ادين”“嘻”
7. L24N1279+1💤	0.71	“新区”“糜”“牛皮”
8. L29N750-1💤	0.70	“明”“udz”“aina”
9. L30N561-1💤	0.70	“川区”“稻”“市新”
10. L29N1181-1🕒	0.70	“棉花糖”“松江”“柿”

Shadow specimen, apart from the ten: L23N8107+1🕒, 0.07.

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 23, neuron 8107, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.71, 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
L23N8107+1🕒 shadow specimen t0-v0.1	0.865	0.69	0.924	1.43	“Selective”“ Short”“ selective”“ am”“选择性”“短期”“ short”“-select”
L23N8107-1 t0-v0.1	0.710	0.75	0.769	1.19	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.