Layer 9 · Neuron 6090 · boosted

L9N6090+1

Word quirks:

“ questions”“ first”“ question”“ usually”“ discussion”“ typically”“ often”“ typical”

💤 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 an AI assistant designed to help users with various tasks. My primary mission is to provide accurate information and support in a helpful manner. If you have specific questions about my capabilities or limitations, please let me know!

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. L12N1180+1🕒	0.93	“ typically”“ question”“ usually”
2. L13N5736+1💤	0.91	“ typically”“ often”“ question”
3. L9N7088-1🕒	0.89	“ typical”“ typically”“ usually”
4. L11N6039-1💤	0.88	“ first”“ usually”“ question”
5. L7N2208-1🕒	0.87	“ typical”“ typically”“ question”
6. L12N962+1🕒	0.87	“ your”“ typically”“ typical”
7. L9N698+1💤	0.84	“ often”“ sometimes”“⏎”
8. L8N4709+1🧪	0.84	“<\|im_end\|>”“ usually”“ without”
9. L7N1299-1🕒	0.84	“ typical”“ my”“?...”
10. L10N6390+1🕒	0.84	“ questions”“ usually”“ **"”

Shadow specimen, apart from the ten: L9N6090-1💤, -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 9, neuron 6090, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.85, top-1 agreement 0.67) 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
L9N6090+1 t0-v0.1	0.847	0.67	0.992	1.37	Top boosted tokens ↑
L9N6090-1💤 shadow specimen t0-v0.1	0.771	0.69	0.985	1.24	“⏎”“ ”“ “”“ __”“ ⏎”“ *”“ '”“ **"”

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.