Layer 11 · Neuron 8713 · suppressed

L11N8713-1

Word quirks:

“ specific”“ specifically”“ Specific”“ specialized”“ detailed”“ generally”“ typically”“ specifics”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

Hello! I am Qwen3.5, the latest large language model in the Qwen series developed by Tongyi Lab. My capabilities include:

Advanced Language Understanding: I can analyze complex reasoning tasks, handle nuanced instructions, and understand context better than previous versions.
Specialized Domain Knowledge: I have been trained on diverse data from multiple domains (such as healthcare, law, and engineering), enabling me to assist with specialized tasks like medical diagnosis or legal analysis.
Multilingual Support: I support fluent interaction in over 100 languages, including Chinese dialects like Cantonese and Minnan

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. L11N2815+1💤	0.95	“ specific”“ specifically”“ usually”
2. L13N5929+1🕒	0.95	“ specific”“ specifically”“ specified”
3. L12N6062+1🕒	0.94	“ specific”“ specifically”“ generally”
4. L11N8717-1🕒	0.94	“ specific”“ specialized”“ Specific”
5. L12N4431+1💤	0.94	“ specific”“ specifics”“ specifically”
6. L12N1323+1🕒	0.94	“ specific”“ specifically”“ Specific”
7. L9N2656-1🕒	0.93	“ specific”“ particular”“ typically”
8. L10N7710-1🕒	0.93	“ specific”“ specifics”“ general”
9. L12N1926+1💤	0.93	“ specific”“ specifics”“ specifically”
10. L8N5448+1💤	0.92	“ specific”“ specifically”“⏎”

Shadow specimen, apart from the ten: L11N8713+1🕒, 0.01.

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 11, neuron 8713, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.74, top-1 agreement 0.66) 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
L11N8713+1🕒 shadow specimen t0-v0.1	1.364	0.53	1.586	1.67	“<\|im_end\|>”“ myself”“ my”“ “”“ ‘”“ _____”“ ‘’”“ carefully”
L11N8713-1 t0-v0.1	0.739	0.66	0.938	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.