Layer 3 · Neuron 5152 · suppressed

L3N5152-1

Word quirks:

“⏎”“ ⏎”“ /”“ !”“ ”“ !!”“ :\”“ -”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am Qwen, an artificial intelligence assistant developed by Alibaba. My capabilities include, but are not limited to, understanding and analyzing questions in various fields, such as science, education, economics, and daily life. I aim to provide high-quality, efficient, and professional support across knowledge and business processes.

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. L5N2449-1🕒	0.87	“ !”“ ⏎⏎”“ ⏎⏎”
2. L3N4387+1🕒	0.87	“ ⏎”“ ⏎⏎”“ !”
3. L9N372-1💤	0.87	“ !”“ '”“⏎”
4. L3N2085+1🕒	0.86	“⏎”“ ⏎”“ !”
5. L3N4949-1🕒	0.86	“⏎”“ -”“ ”
6. L9N1861-1💤	0.85	“⏎”“ _”“ ⏎”
7. L4N598-1🕒	0.85	“ -”“⏎”“ ⏎⏎”
8. L7N1168-1💤	0.85	“⏎”“⏎⏎”“ !”
9. L3N852+1🕒	0.85	“⏎”“ -”“!”
10. L4N4374-1🕒	0.85	“⏎”“ ⏎”“ -”

Shadow specimen, apart from the ten: L3N5152+1💥, -0.18.

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 3, neuron 5152, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L3N5152+1💥 shadow specimen t0-v0.1	3.971	0.26	4.615	2.99	“ disposition”“ distinction”“ divine”“ intellect”“ relation”“ virtue”“ essence”“ Divine”
L3N5152-1 t0-v0.1	1.449	0.53	1.733	1.85	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.