Layer 18 · Neuron 6363 · suppressed

L18N6363-1

Word quirks:

“ here”“ there”“ this”“here”“此处”“ I”“ now”“此为”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I am Qwen3.5, an advanced language model developed by Alibaba I am designed to assist users in completing tasks through text-based interactions. My core mission is to provide high-quality support across a wide range of domains, including:

Specialized capabilities: I support multi-language communication (over 100 languages) and handle complex tasks like code generation/debuging or scientific reasoning.
**I am a specialized assistant focused on delivering reliable, efficient solutions for users' needs.

My primary goal is to empower users with precise information and tools to solve problems effectively. I aim to be a versatile tool for professionals

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. L19N3331+1💤	0.90	“ here”“这里”“here”
2. L18N3333-1🕒	0.90	“ here”“在这里”“ this”
3. L18N5764+1🕒	0.90	“ here”“这里”“ this”
4. L19N6479+1💤	0.89	“ here”“here”“ this”
5. L29N5171-1🕒	0.88	“这里的”“ here”“here”
6. L20N3388+1🕒	0.88	“ here”“ herein”“here”
7. L23N712-1💤	0.86	“ here”“ Here”“here”
8. L22N2942-1🕒	0.85	“ here”“这里”“here”
9. L20N2865+1🕒	0.85	“ here”“ Here”“Here”
10. L31N723-1💥	0.85	“ here”“这里”“here”

Shadow specimen, apart from the ten: L18N6363+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 18, neuron 6363, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.51, top-1 agreement 0.76) 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
L18N6363+1🕒 shadow specimen t0-v0.1	1.574	0.47	1.767	1.40	“ since”“ while”“ although”“ because”“ if”“ not”“既然”“ Since”
L18N6363-1 t0-v0.1	0.507	0.76	0.593	1.09	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.