Layer 17 · Neuron 1687 · suppressed

L17N1687-1

Word quirks:

“ wouldn”“ wanting”“ would”“ before”“ Would”“ perfect”“ needing”“ seeking”

💤 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, and my mission is to be a highly capable, trustworthy, and efficient AI assistant. Here's how I align with this goal:

Helpful: I aim to solve problems collaboratively, such as solving math problems, creating content, or analyzing data before providing answers.
Transparent: I can explain my reasoning process when needed, ensuring clarity and accountability.
Ethical: I prioritize user privacy, adhere to safety guidelines, and maintain confidentiality.
Proactive: I can take initiative in tasks like drafting reports, conducting research, or collaborating on projects.

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. L18N8213+1🕒	0.76	“ needs”“ should”“ desired”
2. L16N2590-1💤	0.75	“ shouldn”“.”“ what”
3. L16N5808+1🕒	0.74	“ would”“ Would”“would”
4. L16N6020-1💤	0.73	“ wanting”“ wanna”“ wanted”
5. L12N7732+1🕒	0.73	“ desired”“ what”“ would”
6. L17N4136-1🕒	0.72	“ before”“ requesting”“ trying”
7. L14N7115+1💤	0.72	“ would”“ ⏎”“ ⏎⏎”
8. L16N8590+1💤	0.72	“ maybe”“ needed”“⏎”
9. L12N6890-1🕒	0.72	“ needed”“ wants”“ needs”
10. L14N1983-1💤	0.72	“ needed”“ could”“ cannot”

Shadow specimen, apart from the ten: L17N1687+1💤, 0.11.

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 17, neuron 1687, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.57, top-1 agreement 0.69) 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
L17N1687+1💤 shadow specimen t0-v0.1	0.607	0.73	0.693	1.10	“ discarded”“ disable”“ disrupted”“ disabled”“ “[”“ disables”“ messed”“ disruption”
L17N1687-1 t0-v0.1	0.565	0.69	0.738	1.25	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.