Layer 17 · Neuron 2629 · suppressed

L17N2629-1

Word quirks:

“ defines”“ exists”“ is”“ doesn”“ qualifies”“ seems”“ constitutes”“ isn”

💤 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, the latest large language model in the Qwen series. I am designed to assist users like you by providing accurate, reliable, and high-quality information, while also helping with complex tasks such as reasoning, coding, and multi-step problem-solving.

My mission is to empower individuals and organizations by enabling them to leverage knowledge and creativity to drive innovation and productivity. This involves:

Enhancing Understanding: Facilitating deeper comprehension of complex topics through clear explanations and nuanced analysis.
Promoting Collaboration: Encouraging constructive dialogue and collaborative efforts across disciplines and cultures.
**Driving

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. L18N3144+1🕒	0.82	“ represents”“ appears”“ constitutes”
2. L16N2852+1🕒	0.80	“ is”“ constitutes”“ represents”
3. L16N4461+1💤	0.79	“ does”“ seems”“ is”
4. L16N6579-1🕒	0.79	“ constitutes”“ qualifies”“ describes”
5. L15N5389+1🕒	0.78	“ doesn”“ is”“ could”
6. L13N3404-1💤	0.77	“ isn”“ doesn”“ is”
7. L16N2529+1🕒	0.77	“ seems”“ feels”“ doesn”
8. L18N420+1🕒	0.77	“ doesn”“ does”“ appears”
9. L15N5044+1🕒	0.77	“ is”“ refers”“ represents”
10. L17N1255-1🕒	0.76	“ isn”“ constitutes”“ doesn”

Shadow specimen, apart from the ten: L17N2629+1🕒, 0.06.

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 2629, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.48, top-1 agreement 0.72) 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
L17N2629+1🕒 shadow specimen t0-v0.1	2.187	0.39	2.332	2.04	“ controlling”“ governing”“ overseeing”“ control”“ aiming”“ regarding”“ controls”“控制”
L17N2629-1 t0-v0.1	0.478	0.72	0.604	1.17	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.