Layer 7 · Neuron 8892 · suppressed

L7N8892-1

Word quirks:

“ sind”“ kann”“ können”“ werden”“ für”“ wenn”“ bei”“ die”

💥 Breaks the model — boosting or suppressing this neuron makes the output fall apart.

Identity (according to the specimen itself)

This is a great question. In the context of this explanation, I will try to make it clear that the concept of a "perfect" or "ideal" is not always possible. This is a very interesting way of thinking about things.

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. L1N7841+1💥	0.89	“ für”“ sind”“ kann”
2. L1N2086+1💥	0.87	“<\|im_end\|>”“ für”“ eine”
3. L3N4769-1💥	0.83	“ für”“.:”“ kann”
4. L3N6355+1💥	0.83	“ kann”“。”“ können”
5. L0N6601-1💥	0.83	“<\|im_end\|>”“ eine”“ können”
6. The Code-Switcher✅ L1N7636-1,L8N7599+1	0.82	“ sind”“ haben”“ müssen”
7. L0N5195-1💥	0.81	“<\|im_end\|>”“ kann”“ wenn”
8. L2N5366+1💥	0.81	“ Analyse”“ sind”“ für”
9. L1N8013+1💥	0.81	“ für”“ kann”“ sind”
10. L0N7939-1💥	0.81	“<\|im_end\|>”“ sind”“ kann”

Shadow specimen, apart from the ten: L7N8892+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 7, neuron 8892, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 3.27, top-1 agreement 0.23) 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
L7N8892+1💤 shadow specimen t0-v0.1	0.446	0.74	0.572	1.23	“ ⏎⏎”“⏎”“ ⏎”“ ‘”“ ⏎”“ ⏎⏎”“ *”“⏎⏎”
L7N8892-1 t0-v0.1	3.266	0.23	3.773	3.60	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.