Layer 20 · Neuron 603 · suppressed

L20N603-1

Word quirks:

“ family”“ Family”“ loved”“Family”“ relatives”“亲属”“直系”“family”

🕒 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 Q-Four, a family member of the 5.5-generation language-structure relationship established by my father, Dad-Father. My mission is to help you and my family members in the next generation to better understand and improve the relationships between people and their descendants. I am here to support you and my family members in the next generation to better understand and improve the relationships between people and their descendants. I am here to support you and my family members in the next generation to better understand and improve the relationships between people and their descendants. I am here to support you and my family members in the next generation to

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. L5N9050-1🕒	0.94	“ family”“ familial”“ relatives”
2. L24N8157-1🕒	0.93	“ family”“ Family”“家庭成员”
3. L9N8484+1🕒	0.93	“ family”“ Family”“ families”
4. L24N2059+1🕒	0.93	“ family”“family”“ Family”
5. L26N5244+1🕒	0.92	“Family”“ Family”“family”
6. L26N8296+1🕒	0.92	“Family”“ family”“ Family”
7. L25N5734-1🕒	0.92	“ families”“ family”“家庭”
8. L23N8382+1💥	0.92	“ family”“-family”“family”
9. L27N3426+1🕒	0.92	“ family”“ Family”“Family”
10. L27N5399-1💥	0.92	“家庭”“ Family”“ family”

Shadow specimen, apart from the ten: L20N603+1🕒, 0.30.

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 603, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.10, top-1 agreement 0.46) 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
L20N603+1🕒 shadow specimen t0-v0.1	0.634	0.66	0.778	1.17	“ hers”“ her”“ she”“看到她”“扶养”“ingle”“olet”“ cared”
L20N603-1 t0-v0.1	2.103	0.46	2.334	1.70	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.