Layer 18 · Neuron 371 · suppressed

L18N371-1

Word quirks:

“ bundle”“ biz”“ replic”“interop”“ distribution”“ sourcing”“ plugged”“ bundles”

🕒 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 Qwen, an advanced AI model developed by Tong. My mission is to serve as a versatile and intelligent partner for you, designed to support various tasks across different industries. Here's how I aim to add value:

My Role & Capabilities

Intelligent Assistant: I can handle complex tasks like content creation, data analysis, and even acting as a research tool for professionals.
Adaptive Support: Whether you need detailed reports, interactive media, or long-form content, I adapt to your needs with precision.
Human-Centric Design: I strive to provide seamless integration into

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. L24N6392+1💤	0.68	“ bundle”“bundle”“ Bundle”
2. L2N8937-1💥	0.68	“ bundle”“<\|im_end\|>”“ bundles”
3. L24N5436-1🕒	0.68	“ bundle”“ bundles”“ Bund”
4. L25N1321-1🕒	0.63	“osz”“ bundle”“.bundle”
5. The Spec Sheet✅ L7N6570+1	0.63	“ bundles”“ your”“ cables”
6. L25N6351+1🕒	0.62	“ package”“ packages”“ bundles”
7. L12N6952-1🕒	0.61	“ packages”“ packaged”“ package”
8. L27N5842+1🕒	0.61	“bundle”“ bundle”“Bundle”
9. L2N4106+1💥	0.61	“ interoper”“ integration”“<\|im_end\|>”
10. L26N2796-1🕒	0.61	“ bundle”“ Bundle”“Bundle”

Shadow specimen, apart from the ten: L18N371+1🕒, -0.14.

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 371, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.67, 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
L18N371+1🕒 shadow specimen t0-v0.1	1.633	0.55	1.913	1.64	“ statements”“ correct”“ Bob”“ statement”“最少”“小王”“ correctly”“小明”
L18N371-1 t0-v0.1	0.666	0.68	0.768	1.18	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.