Layer 4 · Neuron 2558 · suppressed

L4N2558-1

Word quirks:

“⏎”“ ‘”“ ⏎”“ --”“ ⏎”“?\”“ '”“ name”

💤 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 by Alibaba Cloud. My mission is to assist you in any way I can, providing accurate, helpful, and safe information within the context of my capabilities. How can I help you today?

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. L3N3297+1🕒	0.94	“⏎”“<\|im_end\|>”“ ⏎”
2. L7N7139+1💤	0.94	“⏎”“ ⏎”“ ⏎”
3. L3N7100+1💤	0.93	“ ⏎”“ ‘”“ \"”
4. L7N336+1💤	0.93	“⏎”“ ⏎”“ ‘”
5. L7N3204+1💤	0.93	“⏎”“ ‘”“ ⏎”
6. Mirror Neuron✅ L4N468-1	0.93	“⏎”“ ‘”“ ⏎”
7. L7N2629-1💤	0.93	“ \"”“⏎”“ '”
8. L7N1508+1🕒	0.93	“ \"”“ ‘”“ ⏎”
9. L3N6464-1🕒	0.92	“ ⏎”“ ‘”“⏎”
10. L5N7606+1💤	0.92	“⏎”“ '”“ ‘”

Shadow specimen, apart from the ten: L4N2558+1💥, -0.19.

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 4, neuron 2558, 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 0.69) 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
L4N2558+1💥 shadow specimen t0-v0.1	3.510	0.26	4.047	2.80	“ divine”“<\|im_end\|>”“ goddess”“ gods”“ wings”“ grace”“ god”“ himself”
L4N2558-1 t0-v0.1	0.741	0.69	0.934	1.41	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.