Layer 22 · Neuron 4650 · suppressed

L22N4650-1

Word quirks:

“满意”“满意的”“ choosing”“选择”“ approve”“选定”“ choose”“ approving”

🕒 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 Qitixy (also known as Qwixy), a highly intelligent and versatile AI assistant designed to assist with a wide range of tasks. My mission is to provide reliable, accurate, and satisfied assistance by actively engaging in the conversation to understand your needs and offer tailored solutions.

Here's what I aim to deliver:

Trustworthy Confidence: Ensure confidence in every response through rigorous verification and confidence assessment.
Satisfaction Guarantee: Strive for 100% satisfaction by actively engaging in the conversation to understand your needs and offer tailored solutions.
**Active

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. L20N5419+1🕒	0.81	“满足”“选择”“满足了”
2. L22N3893+1🕒	0.81	“选择的”“选择一个”“ chosen”
3. L28N2577+1🕒	0.80	“選択”“选择在”“选择”
4. L19N8994+1💤	0.80	“选择”“ known”“ choose”
5. L23N4505+1💤	0.79	“选了”“选择了”“laud”
6. L22N6647+1🕒	0.79	“选择”“ choice”“ selection”
7. L20N3104+1🕒	0.79	“自由选择”“ choosing”“选择了”
8. L21N5967-1🕒	0.79	“选择一个”“选定”“选择”
9. L19N7741+1🕒	0.78	“ choose”“选择”“ choosing”
10. L23N5547-1🕒	0.78	“选择”“ choice”“ choosing”

Shadow specimen, apart from the ten: L22N4650+1💤, 0.26.

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 22, neuron 4650, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.02, top-1 agreement 0.67) 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
L22N4650+1💤 shadow specimen t0-v0.1	0.996	0.66	1.133	1.46	“由”“ separation”“ separated”“由中国”“罪”“Separated”“由其”“ differentiated”
L22N4650-1 t0-v0.1	1.023	0.67	1.147	1.62	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.