Layer 18 · Neuron 6304 · suppressed

L18N6304-1

Word quirks:

“ expertise”“经验丰富”“ experience”“ capability”“经验”“经验丰富的”“expertise”“擅长”

🕒 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 a large-scale language model developed by the company Tongyi (specifically, I am part of the team behind creating advanced AI capabilities like text generation, code writing, and logical reasoning). My name is Qwen (the model is Qwen3.5).

Mission & Expertise:

Core Strength: I can handle complex tasks like long-form content creation, multi-scenario dialogue, and technical problem-solving (e.g., debugging, reports, and data analysis).
Specialized Skills: I support high-quality text generation, structured data handling, and cross-domain reasoning through advanced

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. Overqualified✅ L25N1825-1	0.87	“ experienced”“ Experienced”“perienced”
2. L20N6817-1🕒	0.81	“ experience”“ Experience”“Experience”
3. L22N8513+1💥	0.81	“经验”“ experience”“的经验”
4. L24N6466-1🕒	0.81	“经验”“的经验”“经验的”
5. L19N2579+1💤	0.81	“ experienced”“ Experienced”“ '$”
6. L24N2559-1🕒	0.80	“ experience”“经验”“ Experience”
7. L28N2079-1🕒	0.80	“ experience”“经验”“experience”
8. L29N2118-1💥	0.80	“经验”“的经验”“经验的”
9. L21N7526+1🕒	0.79	“经验”“ experience”“ Experience”
10. L28N2608+1🕒	0.77	“体验”“ experience”“经验”

Shadow specimen, apart from the ten: L18N6304+1🕒, 0.03.

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 6304, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.99, top-1 agreement 0.70) 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
L18N6304+1🕒 shadow specimen t0-v0.1	0.662	0.68	0.856	1.26	“ choice”“ receiving”“ receive”“ separate”“ chosen”“ choices”“ distinct”“ average”
L18N6304-1 t0-v0.1	0.986	0.70	1.076	1.29	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.