Layer 18 · Neuron 6170 · boosted

L18N6170+1

Word quirks:

“ divides”“venue”“�”“ venue”“ pains”“ dy”“ url”“:__”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am Q*, a model developed by the *Q team. My goal is to help users understand complex concepts, solve problems, and navigate through various challenges. Whether you're dealing with technical issues, managing a project, or just trying to figure out something that's hard to grasp, I'm here to guide you step-by-step.

If you're not sure what to do, feel free to ask me about anything—whether it's related to technology, business, or even something like "how to handle a situation where things aren't going as expected." We can explore the path together! 😌👆

What

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. L2N7677+1💥	0.58	“ venue”“<\|im_end\|>”“ Venue”
2. L23N9145+1🕒	0.57	“ venue”“ ven”“ Ven”
3. L22N2851-1🕒	0.54	“场地”“ venue”“ Venue”
4. L8N6651+1🕒	0.51	“ locations”“ location”“ travel”
5. L20N6619-1🕒	0.48	“ site”“ place”“ sitio”
6. L5N8693+1🕒	0.47	“ hotel”“ banquet”“ hotels”
7. L7N6559-1💤	0.45	“ site”“ response”“ hosting”
8. L8N5725+1🕒	0.45	“ rooms”“ space”“ spaces”
9. L1N3276+1💥	0.44	“<\|im_end\|>”“ site”“ Site”
10. L19N3451+1💤	0.44	“ divide”“ vent”“ dividing”

Shadow specimen, apart from the ten: L18N6170-1💤, -0.06.

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 18, neuron 6170, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L18N6170+1 t0-v0.1	1.029	0.63	1.065	1.29	Top boosted tokens ↑
L18N6170-1💤 shadow specimen t0-v0.1	0.608	0.69	0.766	1.27	“ morning”“ wake”“每天早上”“ wakes”“ waking”“ Morning”“ mornings”“ before”

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.