Layer 5 · Neuron 3999 · suppressed

L5N3999-1

Word quirks:

“ answer”“ question”“ answers”“ questions”“ answering”“ response”“ responses”“ answered”

🕒 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 Qwen3.5, a large language model developed by Alibaba Cloud. My mission is to provide accurate, helpful, and safe responses to your questions. I strive to be as useful as possible while adhering to the highest standards of truthfulness and clarity.

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. L9N6459-1🕒	1.00	“ answer”“ answers”“ responses”
2. Pop Quiz✅ L8N1567+1	0.98	“ answer”“ question”“ answers”
3. L9N2229-1💤	0.97	“ answer”“ answers”“ question”
4. L16N1562-1🕒	0.97	“ answered”“ answer”“ answering”
5. L11N5662-1🕒	0.96	“ answer”“ answers”“ safety”
6. L9N6416-1💤	0.96	“ answer”“ answering”“ question”
7. L9N491+1🕒	0.95	“ response”“ answer”“ responses”
8. L12N6094+1🕒	0.95	“ answer”“ answers”“ unanswered”
9. L14N4564+1💤	0.95	“ answers”“ questions”“ answer”
10. L7N5199+1🕒	0.95	“ answer”“ answering”“ answers”

Shadow specimen, apart from the ten: L5N3999+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 5, neuron 3999, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.91, top-1 agreement 0.49) 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
L5N3999+1💤 shadow specimen t0-v0.1	0.964	0.62	1.173	1.63	“⏎”“ your”“ —”“ ——”“ reference”“ '”“ yourself”“ ⏎”
L5N3999-1 t0-v0.1	1.911	0.49	2.313	1.71	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.