Layer 30 · Neuron 8043 · suppressed

L30N8043-1

Word quirks:

“ivit”“看不”“UIT”“eting”“细”“гат”“满”“ect”

💤 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 Qwens, a large-scale language model developed by Tencent. My primary mission is to assist users in solving problems across a wide range of domains—whether that's answering factual questions, creative writing, logical reasoning, coding, or even deep diving into complex topics. I strive to provide accurate, efficient, and human-like interactions while adhering to ethical and safety guidelines.

Think of me as a versatile digital companion designed to:

Accelerate learning (e.g., explaining concepts clearly)
Spark creativity (e.g., crafting stories, poems

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. L29N1981-1💤	0.67	“ect”“涩”“iday”
2. L21N4731-1🕒	0.66	“膨”“abu”“igits”
3. L25N5844+1🕒	0.65	“精”“贮”“耳目”
4. L30N6911+1🕒	0.65	“利”“egger”“返”
5. L29N8366+1🕒	0.65	“itas”“mit”“pit”
6. L28N7743+1💤	0.65	“eto”“看”“emar”
7. L30N4309-1💤	0.65	“油”“达”“atus”
8. L29N143-1💤	0.65	“atern”“兴”“统一”
9. L28N3970+1💤	0.65	“attel”“融”“望”
10. L29N2184-1🕒	0.65	“fty”“egral”“etter”

Shadow specimen, apart from the ten: L30N8043+1💤, 0.39.

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 30, neuron 8043, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.38, top-1 agreement 0.76) 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
L30N8043+1💤 shadow specimen t0-v0.1	0.328	0.82	0.346	1.70	“知”“明”“base”“mina”“mine”“ Herm”“len”“基地”
L30N8043-1 t0-v0.1	0.383	0.76	0.488	1.64	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.