Layer 20 · Neuron 2673 · suppressed

L20N2673-1

Word quirks:

“ horizontal”“ vertical”“的水平”“ Vertical”“ Horizontal”

🕒 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 Qwaver, a specialized AI assistant designed to help you with tasks ranging from creative writing to technical problem-solving. My mission is to provide you with efficient, accurate, and engaging support by breaking down complex problems into manageable steps and offering clear solutions. Whether you need help with a specific project, a question, or just a bit of assistance, I'm here to help! 😊

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. L30N6343+1🕒	0.98	“horizontal”“ horizontal”“ Horizontal”“ vertical”“Horizontal”
2. L28N6684-1🕒	0.97	“Horizontal”“ horizontal”“ horizontally”“orizontal”“ vertical”
3. L25N2901-1🕒	0.96	“ horizontal”“ Horizontal”“横向”“ horizontally”“Horizontal”
4. L26N2615-1🕒	0.96	“ vertical”“vertical”“ Vertical”“Vertical”“ horizontal”
5. L24N4777+1💥	0.95	“ horizontal”“ Horizontal”“水平”“Horizontal”“ vertical”
6. L0N7035+1💥	0.95	“ horiz”“vertical”“horizontal”“ horizontal”“ vertical”
7. L24N3229-1💤	0.95	“ horizontal”“Horizontal”“ Horizontal”“ Hor”“-horizontal”
8. L26N5760+1🕒	0.93	“ horizontal”“ Horizontal”“horizontal”“Horizontal”“ гориз”
9. L3N3816-1🕒	0.93	“ vertical”“ horizontal”“ vertically”“-vertical”“ horiz”
10. L19N6197-1💤	0.93	“ horizontal”“ horizontally”“horizontal”“ Horizontal”“横向”

Shadow specimen, apart from the ten: L20N2673+1🕒, 0.12.

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 2673, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.78, top-1 agreement 0.68) 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
L20N2673+1🕒 shadow specimen t0-v0.1	0.655	0.75	0.730	1.26	“ older”“acen”“ exempt”“ışı”“กำ”
L20N2673-1 t0-v0.1	0.777	0.68	0.892	1.26	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.