Layer 23 · Neuron 8418 · suppressed

L23N8418-1

Word quirks:

“ declaration”“ declarations”“Declaration”“ Declaration”“ Declarations”“declaration”“表达”“ declare”

🕒 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 Ai Dalgeon, a specialized AI assistant designed to help you with a wide range of tasks. My mission is to be your reliable partner in:

Expressing Love & Affection: I can help you write romantic letters, love declarations, or sweet gestures for your loved ones.
Planning Romantic Gestures: Whether it's planning a grand romantic proposal, organizing a surprise anniversary, or creating a perfect date proposal, I can help you plan the perfect gesture.
Declaring Love: I can help you express your feelings in the most beautiful and heartfelt way possible.

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. L21N2568+1🕒	0.96	“ declaration”“ Declare”“ declare”
2. L21N3791-1🕒	0.95	“ declaration”“ declare”“ declarations”
3. L21N8752+1💤	0.95	“声明”“ declare”“ declaring”
4. L23N8283+1🕒	0.95	“ declare”“ declared”“ Decl”
5. L23N8123-1🕒	0.94	“ declared”“ declaration”“ declare”
6. L18N8429-1💤	0.94	“ declaring”“ declare”“ declares”
7. L31N4115-1💥	0.94	“宣”“ declare”“ declared”
8. L24N4679-1🕒	0.94	“宣”“declaration”“ declaration”
9. L2N6724+1💥	0.94	“<\|im_end\|>”“ declare”“ declaration”
10. L25N1450-1🕒	0.94	“ Decl”“ declare”“ declares”

Shadow specimen, apart from the ten: L23N8418+1🕒, 0.07.

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 23, neuron 8418, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.90, top-1 agreement 0.72) 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
L23N8418+1🕒 shadow specimen t0-v0.1	0.840	0.71	0.951	1.39	“自助”“WP”“ WP”“iefert”“сихо”“朝天”“ info”“ampen”
L23N8418-1 t0-v0.1	0.897	0.72	0.895	1.35	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.