Layer 18 · Neuron 4685 · suppressed

L18N4685-1

Word quirks:

“,and”“ ⏎”“ ⏎”“–and”“—and”“⏎”“ ⏎”“ และการ”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

The reply didn't hold together

And I am Qwen, and my mission is to empower people, and to make life easier, and to make life better. And I am here, and I am ready, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I am here, and I

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. L18N282+1💤	0.86	“ and”“ ⏎”“and”
2. L8N1549-1🕒	0.85	“ and”“ &”“ AND”
3. L17N5478-1💤	0.83	“ and”“ AND”“ ⏎”
4. L17N8685+1🕒	0.83	“⏎”“ and”“，“”
5. L18N6500+1🕒	0.82	“,and”“ and”“,”
6. L16N5671+1🕒	0.82	“⏎”“—and”“ ⏎”
7. L11N6843+1🕒	0.80	“ and”“⏎”“ ⏎”
8. L29N1282-1💥	0.80	“—and”“,and”“–and”
9. L21N5203+1🕒	0.80	“—and”“,and”“ and”
10. L20N685+1💤	0.80	“,”“，”“⏎”

Shadow specimen, apart from the ten: L18N4685+1💤, 0.08.

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 18, neuron 4685, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L18N4685+1💤 shadow specimen t0-v0.1	0.401	0.81	0.495	1.15	“":”“\":”“:”“:”“”:”“':”“*:”“}:”
L18N4685-1 t0-v0.1	1.119	0.59	1.221	1.15	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.