Layer 18 · Neuron 7274 · suppressed

L18N7274-1

Word quirks:

“ aligned”“ alignment”“ axis”“ aline”“ align”“ alignments”“ direction”“ Alignment”

🕒 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, also known as Wangqi (in the direction of the original Chinese), and I am a large language model developed by Alibaba Cloud's AI Lab. My mission is to assist users in their tasks by providing high-quality responses that align with the following principles:

Authenticity: I strive to deliver accurate, relevant information aligned with the task at hand.
Efficiency: I aim to complete tasks within the plane of the current task, ensuring alignment with the user's intent.
Transparency: I will clearly state my capabilities and

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. L20N4796-1🕒	0.97	“ alignment”“ aligned”“ align”
2. L13N7979+1🕒	0.96	“ align”“ aligned”“ alignment”
3. L21N4752+1💤	0.96	“ align”“align”“ aligned”
4. L23N1758+1🕒	0.95	“ alignment”“ aligned”“ Alignment”
5. L24N1610+1💤	0.95	“ aligned”“ align”“ Align”
6. L7N2914-1🕒	0.95	“ alignment”“ align”“ aligned”
7. L23N1965-1💤	0.95	“ align”“ alignment”“ aligned”
8. L2N5739-1💥	0.95	“ alignment”“ alignments”“ aligned”
9. L6N8494+1🕒	0.94	“ alignment”“ aligned”“ align”
10. L24N3575-1🕒	0.94	“ alignment”“ Alignment”“alignment”

Shadow specimen, apart from the ten: L18N7274+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 7274, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.92, top-1 agreement 0.64) 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
L18N7274+1🕒 shadow specimen t0-v0.1	0.663	0.71	0.627	1.17	“—we”“ elasticity”“ free”“ymbol”“free”“price”“自由的”“Free”
L18N7274-1 t0-v0.1	0.917	0.64	1.112	1.47	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.