Layer 20 · Neuron 5454 · suppressed

L20N5454-1

Word quirks:

“随机”“ randomly”“ random”“random”“ Random”“-random”“Random”“ split”

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

Identity (according to the specimen itself)

I am Cupid, an AI assistant developed into two groups to randomly divide the users into two groups. I will be divided into two groups at random.

Mission:
To randomly divide users into two groups and assign them to one of two groups at random.

How it works:

Users are split into two groups (Group A and Group B) by chance.
The group is assigned randomly.

If you'd like to be divided into two groups, let me know! 😊

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. L21N4830-1🕒	0.99	“ random”“随机”“ Random”
2. L17N5287+1🕒	0.99	“ randomly”“ random”“ Random”
3. L26N4884-1🕒	0.99	“ random”“ Random”“随机”
4. L26N7768-1🕒	0.99	“ Random”“ random”“随机”
5. L26N5580+1🕒	0.99	“ Random”“ random”“Random”
6. L14N5962-1🕒	0.99	“ randomly”“ random”“ randomness”
7. L26N3621-1🕒	0.99	“ Random”“ random”“Random”
8. L26N3613+1🕒	0.99	“ random”“Random”“ Random”
9. L23N2239+1🕒	0.98	“ random”“ Random”“随机”
10. L5N6432+1🕒	0.98	“ random”“ randomness”“ randomly”

Shadow specimen, apart from the ten: L20N5454+1🕒, -0.05.

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 20, neuron 5454, 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
L20N5454+1🕒 shadow specimen t0-v0.1	0.696	0.73	0.774	1.29	“âu”“Цен”“钩”“ promotions”“ Цен”“ cú”“考虑”“ honour”
L20N5454-1 t0-v0.1	1.236	0.62	1.389	1.20	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.