Layer 20 · Neuron 7333 · boosted

L20N7333+1

Word quirks:

“secured”“ secured”“porter”“room”“ setups”“ Establishment”“ effected”“ stores”

🕒 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 Qianxin, a large-scale Chinese model of a comprehensive Chinese assistant. My mission is to provide helpful, accurate, and safe information across a wide range of domains, including but not limited to:

Knowledgeable Answers: I can answer questions on history, science, languages, etc., with precise information.
Practical Assistance: I can help with daily tasks like writing content, summarizing text, or solving problems through reasoning.
Safe & Reliable Information: I prioritize accurate information and avoid harmful content, ensuring that the information provided is safe and reliable.
**Efficient & Helpful

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. L7N7925-1🕒	0.67	“ facility”“ property”“ housing”
2. L30N6258-1💥	0.63	“ establishment”“ Establishment”“ establishments”
3. L5N8101-1🕒	0.60	“ storage”“ secure”“ securely”
4. L6N9086+1🕒	0.59	“ facilities”“ available”“ equipment”
5. L21N558-1🕒	0.59	“ secured”“ held”“ exposed”
6. L9N8039-1🕒	0.59	“ access”“ security”“ opened”
7. L10N8959-1🕒	0.59	“ facility”“ company”“ customers”
8. L9N8573-1🕒	0.59	“ detention”“ accommodation”“ housing”
9. L24N3278-1🕒	0.59	“ establishment”“establish”“ establishments”
10. L11N7212-1🕒	0.58	“ opened”“ stores”“ facilities”

Shadow specimen, apart from the ten: L20N7333-1🕒, 0.16.

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 20, neuron 7333, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.79, top-1 agreement 0.67) 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
L20N7333+1 t0-v0.1	0.793	0.67	1.026	1.26	Top boosted tokens ↑
L20N7333-1🕒 shadow specimen t0-v0.1	0.928	0.61	0.967	1.41	“ cards”“cards”“Cards”“ Cards”“卡片”“ coins”“ card”“执业”

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.