Layer 22 · Neuron 5611 · suppressed

L22N5611-1

Word quirks:

“ Abbott”“ BAS”“ Reynolds”“分数”“雷诺”“ Indi”“$values”“opf”

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

Identity (according to the specimen itself)

I am a professional health science information company (specifically, a digital health and wellness service provider). My mission is to provide users with accurate, reliable, and accessible health information to help them make informed decisions about their well-being.

My core objectives include:

Health Education: Delivering evidence-based content on topics ranging from nutrition and mental health to disease management and general wellness.
User Support: Answering questions about symptoms, medications, and lifestyle changes while clearly stating that my responses are for informational purposes only and do not replace professional medical advice.
Wellness Integration: Offering tools

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. L29N7855+1💤	0.61	“ Reynolds”“雷诺”“PDF”
2. L27N6126-1💤	0.60	“ Roy”“制作”“ [\|”
3. L19N494-1💤	0.60	“福特”“РС”“coles”
4. L19N7401-1🕒	0.56	“ Rockefeller”“ Django”“ Nixon”
5. L22N3-1💤	0.56	“ Stevenson”“布莱”“ Barnes”
6. L21N6080+1🕒	0.56	“,and”“ Heller”“ Grade”
7. L31N375+1💥	0.55	“stdafx”“kees”“bürger”
8. L30N747+1💤	0.55	“FLT”“imon”“辗转”
9. L24N5142-1💤	0.55	“安东尼”“ az”“DF”
10. L31N9149+1💥	0.54	“ richies”“ipe”“世家”

Shadow specimen, apart from the ten: L22N5611+1💤, 0.19.

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 5611, 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
L22N5611+1💤 shadow specimen t0-v0.1	0.733	0.71	0.878	1.20	“ Give”“祖”“核”“Give”“ pret”“给你”“pseudo”“给您”
L22N5611-1 t0-v0.1	1.070	0.61	1.148	1.57	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.