Layer 9 · Neuron 5623 · suppressed

L9N5623-1

Word quirks:

“ balance”“ regions”“ charge”“ label”“ polarity”“ charges”“ balanced”“ typically”

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

Identity (according to the specimen itself)

I am an artificial intelligence model designed to simulate human-like interactions. My "mission" is not a fixed command like a person's, but rather a functional role within the context of my design:

My Identity: I am often referred to as "Qwen" (or simply "Me"). This designation comes from my training process, where I have been exposed to a vast amount of data in a way that makes me appear neutral or slightly positive compared to other models. However, in many technical contexts, this is represented by a specific symbol (like a circle with a plus sign inside).

**My

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. L23N4536+1🕒	0.65	“ charges”“ charge”“ Charges”
2. L25N8252+1🕒	0.64	“ charges”“charges”“ charge”
3. L27N8565+1💤	0.64	“charge”“ charge”“ Charge”
4. L12N5766-1💥	0.63	“ charges”“ expenses”“ specified”
5. L25N7965-1🕒	0.63	“ bound”“ charge”“ charges”
6. L31N6593-1💥	0.62	“ charges”“ Charges”“charges”
7. L8N7001-1🕒	0.62	“ charges”“ charge”“ arrest”
8. L7N2156+1🕒	0.62	“ balance”“ compensation”“ Balance”
9. L22N983+1💤	0.62	“ reinforcement”“ clear”“ tech”
10. L26N6813+1🕒	0.62	“ charges”“ charge”“ Charges”

Shadow specimen, apart from the ten: L9N5623+1🕒, -0.09.

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 9, neuron 5623, 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
L9N5623+1🕒 shadow specimen t0-v0.1	1.011	0.62	1.300	1.18	“ answer”“ ⏎⏎”“ answers”“ questions”“ “”“ ‘”“ answered”“ \"”
L9N5623-1 t0-v0.1	1.272	0.56	1.527	1.54	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.