Layer 16 · Neuron 3139 · suppressed

L16N3139-1

Word quirks:

“ similarly”“ similar”“ nearby”“ corresponding”“ counterparts”“ comparable”“ same”“ analogous”

💤 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 from Alibaba Cloud. My mission is to empower users by providing high-quality, context-aware assistance across a wide range of domains—whether it's solving complex problems, analyzing data, or helping with creative projects. I aim to make the world more accessible and impactful through advanced capabilities like multi-modal understanding, logical reasoning, and deep domain expertise. Together, we can tackle challenges in education, healthcare, science, and beyond! How can I help you today?

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. L16N3999-1🕒	0.96	“ similar”“ same”“ corresponds”
2. L12N7579+1🕒	0.96	“ similar”“ analogous”“ similarly”
3. Stage Directions✅ L15N6354+1	0.96	“ similar”“ analogous”“ equivalent”
4. L14N4051+1💤	0.96	“ similarly”“ corresponding”“ similar”
5. L13N4635+1🕒	0.95	“ similar”“ similarly”“ analogous”
6. L13N1342-1🕒	0.95	“ similar”“ similarly”“ comparable”
7. L15N4728-1💤	0.94	“ similar”“ similarly”“ same”
8. L17N7874+1🕒	0.94	“ similar”“ similarly”“ analogous”
9. L14N3338-1💤	0.94	“ analogous”“ similar”“ counterpart”
10. L11N4495-1🕒	0.94	“ similar”“ similarly”“ analogous”

Shadow specimen, apart from the ten: L16N3139+1💤, 0.08.

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 16, neuron 3139, 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 0.71) 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
L16N3139+1💤 shadow specimen t0-v0.1	0.543	0.68	0.692	1.14	“ after”“ initially”“ if”“ when”“ first”“ before”“ prior”“ now”
L16N3139-1 t0-v0.1	0.568	0.71	0.738	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.