Layer 7 · Neuron 7043 · suppressed

L7N7043-1

Word quirks:

“ special”“ specific”“ specifically”“ specially”“ standard”“ precise”“ identification”“ defining”

💤 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, a large language model developed by Alibaba. My mission is to help users solve complex problems, generate high-quality content, and provide personalized support across various domains such as education, healthcare, finance, and technology. By leveraging advanced natural language processing capabilities, I can assist individuals in achieving their goals more efficiently and effectively. Whether it's drafting a research paper, creating a business plan, or offering tailored advice, my role is to deliver exceptional results that make a meaningful difference in people's lives.

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. L8N9000-1💤	0.91	“ specific”“ special”“ specifically”
2. L7N3773-1🕒	0.90	“ precisely”“ specially”“ specifically”
3. L9N6976+1🕒	0.89	“ specific”“ specifically”“ specially”
4. L10N7649+1🕒	0.89	“ specific”“ specifically”“ specialized”
5. L18N4369-1💤	0.89	“ specific”“ specifically”“ particular”
6. L9N5927+1🕒	0.89	“ specific”“ specially”“ particular”
7. L8N1726+1💤	0.89	“ special”“ specially”“ specific”
8. L8N5170-1🕒	0.88	“ specific”“ specifics”“ Specific”
9. L11N8717-1🕒	0.88	“ specific”“ specialized”“ Specific”
10. L9N5622+1🕒	0.88	“ specific”“ specifically”“ special”

Shadow specimen, apart from the ten: L7N7043+1🕒, 0.09.

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 7, neuron 7043, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.93, top-1 agreement 0.59) 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
L7N7043+1🕒 shadow specimen t0-v0.1	1.617	0.48	1.989	1.83	“ name”“ naming”“ Name”“ implying”“"!”“ \"”“ referring”“<\|im_end\|>”
L7N7043-1 t0-v0.1	0.931	0.59	1.049	1.45	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.