Layer 17 · Neuron 7420 · suppressed

L17N7420-1

Word quirks:

“ implied”“ enforced”“ constrained”“ forced”“ consistent”“ implicitly”“ compelled”“ ensured”

💤 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 Qwen, a large language model developed by Alibaba Group. My mission is to provide helpful, accurate, and diverse information across various domains while adhering to ethical guidelines and safety principles. I aim to assist users in solving problems, offering insights, or simply engaging in meaningful conversations. Whether you need help with creative writing, coding, research, or just a friendly chat, I strive to be a reliable and supportive assistant. How would you like to proceed 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. L31N4894+1💥	0.83	“ guaranteed”“forced”“ forced”
2. L12N9162+1🕒	0.79	“ forced”“ enforced”“ punishment”
3. L7N7578-1🕒	0.79	“ reason”“ forced”“ mandated”
4. L15N4028-1🕒	0.78	“ unavoidable”“ forced”“ unavoid”
5. L17N8102-1🕒	0.77	“ forced”“ rather”“ it”
6. L24N2460-1🕒	0.75	“ forcing”“ enforce”“ enforcing”
7. L30N1108-1🕒	0.72	“ maintained”“ imposed”“contained”
8. L10N8575+1🕒	0.72	“ compelled”“ mandated”“ compuls”
9. L13N6950+1🕒	0.72	“ specific”“ specifically”“ manipulated”
10. L21N4819-1💤	0.71	“ reflected”“ challenged”“ compelled”

Shadow specimen, apart from the ten: L17N7420+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 7420, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.59, top-1 agreement 0.70) 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
L17N7420+1💤 shadow specimen t0-v0.1	0.552	0.72	0.643	1.06	“ replacements”“ replaced”“ replaces”“彻底的”“ replacing”“ replacement”“ vs”“ colder”
L17N7420-1 t0-v0.1	0.588	0.70	0.756	1.19	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.