Layer 6 · Neuron 6629 · suppressed

L6N6629-1

Word quirks:

“ rooms”“ floors”“ halls”“ hall”“ courtyard”“ towers”“ houses”“ floor”

🕒 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 an AI assistant in the city of Shanghai. My mission is to provide support services for visitors and residents, including information desks, rest areas, and guided tours. How can I assist you today?

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. L5N8726+1🕒	0.90	“ floors”“ halls”“ hall”
2. L8N6303+1🕒	0.89	“ floors”“ rooms”“ floor”
3. L5N3995-1🕒	0.88	“ halls”“ houses”“board”
4. L27N1205+1🕒	0.85	“ hall”“ buildings”“ building”
5. L14N5860+1🕒	0.85	“ floors”“ houses”“ pane”
6. L3N8023+1💥	0.84	“ pav”“ halls”“ hall”
7. The Architect✅ L6N7374-1	0.84	“ floors”“ rooms”“ ceilings”
8. L5N7329-1🕒	0.84	“ rooms”“ spaces”“ houses”
9. L7N6272+1🕒	0.83	“ rooms”“ apartment”“ room”
10. L4N5205+1💥	0.83	“ space”“ lobby”“ floors”

Shadow specimen, apart from the ten: L6N6629+1💤, -0.13.

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 6, neuron 6629, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.07, top-1 agreement 0.47) 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
L6N6629+1💤 shadow specimen t0-v0.1	0.604	0.69	0.719	1.27	“⏎”“ _____”“ **"”“ digital”“⏎⏎”“ internet”“ ____”“ ______”
L6N6629-1 t0-v0.1	2.067	0.47	2.459	1.84	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.