Layer 23 · Neuron 5361 · boosted

L23N5361+1

Word quirks:

“XR”“华晨”“功”“成功”“ المخ”“育成”“出世”“买进”

💤 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 Alexandander, a specialized AI assistant designed to provide precise, safe, and helpful responses. My core mission is to serve as a reliable partner in learning, problem-solving, and creative endeavors by:

Answering Questions – From complex research topics to everyday curiosities, I aim for accuracy and clarity.
Solving Problems – Whether it's coding challenges, mathematical proofs, or strategic planning, I offer step-by-step reasoning.
Creating Content – I can write stories, scripts, poems, or even develop interactive scenarios tailored to your needs.
**

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. L29N1999-1🕒	0.82	“成”“成分”“成本”
2. L30N4433+1💤	0.81	“修”“成功”“萧”
3. L20N460-1🕒	0.80	“armo”“看到”“嘉年华”
4. L29N1921+1🕒	0.80	“原”“沙漠”“成功”
5. L30N2983-1💤	0.80	“成功”“行”“功用”
6. L29N3971-1🕒	0.80	“能”“可以”“义”
7. L31N873+1💥	0.79	“录得”“地被”“化成”
8. L31N7658-1💥	0.79	“入”“结果”“策略”
9. L31N5622+1💥	0.79	“干脆”“文献”“女神”
10. L31N8265-1💥	0.79	“完成”“成”“知”

Shadow specimen, apart from the ten: L23N5361-1💤, 0.23.

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 5361, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.78, top-1 agreement 0.71) 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
L23N5361+1 t0-v0.1	0.781	0.71	0.905	1.29	Top boosted tokens ↑
L23N5361-1💤 shadow specimen t0-v0.1	0.792	0.71	0.807	1.32	“分解”“ decomposition”“ decom”“ Decom”“拆解”“Dec”“ Decomp”“ Thermal”

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.