Layer 27 · Neuron 8547 · boosted

L27N8547+1

Word quirks:

“招”“onn”“europ”“萝”“uko”“uyo”“-du”“anzia”

💤 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 Qozi, an AI model developed by Zhangzhou Zhanzuo Technology Co., Ltd. (also known as Zhanzuo AI or Zhanzuo AI Lab). My mission is to empower users with intelligent, ethical, and context-aware solutions across diverse domains such as education, healthcare, finance, and daily life. I strive to:

Resolve complex problems with precision and creativity,
Promote fairness and transparency in AI interactions,
Adapt to evolving needs through continuous learning and user feedback,
**Prior

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. L31N2455+1💥	0.71	“�”“uchos”“ottom”
2. L30N7147-1💤	0.71	“ugo”“南国”“anko”
3. L31N5819+1💥	0.70	“chio”“ро”“昆”
4. L25N9210-1💤	0.70	“萝”“unga”“.ir”
5. L25N4752+1💤	0.69	“lod”“่ว”“alos”
6. L21N3324-1💤	0.68	“ору”“anto”“江河”
7. L30N2537-1💤	0.68	“ uğ”“大招”“enko”
8. L28N3425-1💤	0.68	“DU”“�”“orro”
9. L30N4438+1💤	0.68	“ué”“otra”“ié”
10. L28N1802-1💤	0.67	“phy”“oi”“迫”

Shadow specimen, apart from the ten: L27N8547-1🕒, 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 27, neuron 8547, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.37, top-1 agreement 0.82) 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
L27N8547+1 t0-v0.1	0.372	0.82	0.405	1.27	Top boosted tokens ↑
L27N8547-1🕒 shadow specimen t0-v0.1	0.525	0.72	0.560	1.23	“ head”“头”“-head”“head”“ heads”“headed”“的头”“ Head”

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.