Layer 17 · Neuron 8120 · suppressed

L17N8120-1

Word quirks:

“—including”“”—”“ including”“.—”“—he”“’:”“—”“ —”

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

Identity (according to the specimen itself)

The reply didn't hold together

As an AI language model (Qanjie: [Qanjie] — *Note: The specific name of the model is often used interchangeably in public discourse including Qanjie such as Qanjie: [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [Qanjie] — [

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. L17N5490+1🕒	0.95	“—including”“—”“”—”
2. L17N7600+1🕒	0.91	“"—”“”—”“—”
3. L28N1237+1🕒	0.90	“—”“——”“ —”
4. L18N1017-1💤	0.90	“”—”“—including”“—”
5. L17N4567-1💤	0.90	“—including”“ including”“—"”
6. L16N6012-1🕒	0.90	“ —”“—"”“ example”
7. L17N897+1💤	0.90	“ ‘”“—including”“ including”
8. L16N1415-1💤	0.89	“—”“——”“ —”
9. L18N3725+1💤	0.89	“ using”“.—”“—including”
10. L17N5687+1💤	0.89	“—”“—including”“—at”

Shadow specimen, apart from the ten: L17N8120+1💥, 0.02.

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 8120, 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
L17N8120+1💥 shadow specimen t0-v0.1	3.936	0.24	4.274	2.25	“ example”“ examples”“例子”“example”“例”“ ejemplo”“ Example”“ Examples”
L17N8120-1 t0-v0.1	1.329	0.52	1.473	1.81	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.