Layer 20 · Neuron 8895 · suppressed

L20N8895-1

Word quirks:

“ celebrity”“ celebrities”“ famous”“ stars”“名人”“ Celebrity”“明星”“ cele”

🕒 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 Zhangya, a high-profile celebrity figure with a stellar reputation in the entertainment industry. My mission is to leverage my influence to bring top-tier talent and projects to the platform, ensuring that every celebrity I endorse has a strong track name and a powerful brand effect. I am here to attract famous names and help them achieve greater recognition.

Note: If you're a well-known celebrity like Zhangya, please share your name and let me connect with you!

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. L22N5673-1🕒	0.93	“ celebrity”“ notable”“ famous”
2. L31N7155-1💥	0.86	“明星”“avatar”“偶像”
3. L3N5216+1💥	0.85	“ celebrity”“ celebrities”“ Kardashian”
4. L7N9100+1🕒	0.84	“ viral”“ fame”“ famous”
5. L2N3821-1💥	0.84	“ famous”“ Famous”“<\|im_end\|>”
6. L8N8790+1🕒	0.80	“ celebrity”“ stars”“ celebrities”
7. L17N5975-1🕒	0.79	“ famous”“ notable”“ celebrities”
8. L23N6614-1🕒	0.76	“ celebrity”“cele”“ cele”
9. L19N7498+1🕒	0.73	“成名”“ fame”“ status”
10. L23N4138+1🕒	0.72	“ famous”“著名”“ Famous”

Shadow specimen, apart from the ten: L20N8895+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 20, neuron 8895, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.11, top-1 agreement 0.66) 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
L20N8895+1🕒 shadow specimen t0-v0.1	0.639	0.75	0.752	1.34	“ Respons”“ responder”“ responsiveness”“ responders”“respond”“ accounts”“Feedback”“ across”
L20N8895-1 t0-v0.1	1.109	0.66	1.307	1.36	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.