Layer 17 · Neuron 8092 · boosted

L17N8092+1

Word quirks:

“ relationship”“ relationships”“ Relationships”“关系”“Relationship”“ Relationship”“relationship”“ marriage”

🕒 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)

We are a close relationship, so let's talk about ourselves. We are a young adult with a gentle heart, known for being kind and caring. We have a deep affection for our partner, and we share a loving bond that has brought them together. They are often described as the "beauty" of their life, and they have a reputation for being popular and well-liked in social circles.

In our relationship, they are often the one who takes care of others, making them a responsible and caring individual. They have a reputation for being popular and well-liked, and they are often the one who gets along well with

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. L25N6518-1💥	0.99	“Relationship”“ Relationships”“ relationships”
2. L0N8863-1💥	0.98	“relationship”“Relationship”“ Relationship”
3. L23N7303-1🕒	0.98	“ relationship”“ Relationship”“的关系”
4. L20N5295+1🕒	0.98	“ relationships”“ Relationships”“relationships”
5. L25N4333+1💥	0.98	“关系”“ relationships”“Relationship”
6. L23N2806+1🕒	0.98	“关系”“ relationships”“ Relationship”
7. L19N5399+1💤	0.98	“ relationship”“关系”“ relationships”
8. Token Cupid✅ L21N6298+1	0.97	“ relationships”“ Relationships”“relationships”
9. L25N6654+1🕒	0.97	“ relationship”“关系”“Relationship”
10. L27N2020-1💤	0.97	“关系”“Relationship”“ Relationship”

Shadow specimen, apart from the ten: L17N8092-1🕒, 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 17, neuron 8092, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.88, top-1 agreement 0.49) 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
L17N8092+1 t0-v0.1	1.885	0.49	2.149	1.51	Top boosted tokens ↑
L17N8092-1🕒 shadow specimen t0-v0.1	0.865	0.63	0.960	1.32	“ automatic”“ automated”“ browsers”“ network”“ computers”“ servers”“ nodes”“ networks”

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.