Layer 16 · Neuron 7077 · suppressed

L16N7077-1

Word quirks:

“ choice”“ chosen”“ initially”“ selected”“ choices”“ selectively”“ choosing”“ preferred”

💤 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 trained with a focus on advanced capabilities across multiple domains. My training leverages high-quality data from diverse sources, ensuring broad coverage and adaptability to various scenarios. I also incorporate extensive knowledge updates to maintain relevance and accuracy.

My mission is to:

Prioritize safety: Adhere to strict guidelines to prevent harm, including avoiding sensitive topics and protecting user privacy.
Enhance global reach: Support over 100 languages, enabling widespread adoption across regions.
Optimize performance: Balance speed and efficiency for scalability in large

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. L8N3169+1🕒	0.94	“ choice”“ selection”“ selections”
2. L16N6961+1🕒	0.93	“ choice”“ only”“ choosing”
3. L15N7581+1🕒	0.93	“ chosen”“ selected”“ choice”
4. L7N8600-1🕒	0.93	“ choice”“ choosing”“ chosen”
5. L12N9078-1💤	0.93	“ choice”“ choices”“ selection”
6. L16N4798+1💤	0.92	“ choice”“ chosen”“ choosing”
7. L11N6339-1🕒	0.92	“ choice”“ choose”“ choosing”
8. L12N7372+1💤	0.92	“ choice”“ choices”“ choosing”
9. L16N973-1💤	0.91	“ choosing”“ choice”“ chosen”
10. L16N6441+1💤	0.91	“ choice”“ chosen”“ choices”

Shadow specimen, apart from the ten: L16N7077+1🕒, 0.03.

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 7077, 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 0.80) 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
L16N7077+1🕒 shadow specimen t0-v0.1	1.244	0.57	1.541	1.59	“ representation”“ indicating”“ implying”“ referring”“ means”“ indications”“ representing”“ indication”
L16N7077-1 t0-v0.1	0.484	0.80	0.615	1.20	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.