Layer 28 · Neuron 5894 · boosted

L28N5894+1

Word quirks:

“摆”“眼”“wort”“察”“玉龙”“landers”“ Bieber”“ entreg”

💤 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 Q3, a large-scale language model developed by Dense-Cell. My primary mission is to assist users in solving complex problems across diverse domains such as science, technology, culture, and daily life. I aim to provide accurate, creative, and actionable solutions while maintaining high standards of safety and ethical conduct. Whether you need detailed analysis, code generation, or creative writing, I strive to deliver clear and reliable support tailored to your needs. How can I assist 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. L29N8087+1💤	0.76	“ Pomp”“dı”“维克”
2. L26N8698+1🕒	0.76	“莽”“勃”“odz”
3. L31N2229-1💥	0.76	“木偶”“木瓜”“天庭”
4. L31N3317+1💥	0.76	“жда”“稽”“摆手”
5. L31N5324-1💥	0.76	“神龙”“苓”“baikan”
6. L29N4792+1🕒	0.75	“otten”“fil”“千”
7. L28N8414-1💤	0.75	“陈”“пара”“爱尔”
8. L28N5934-1💤	0.75	“usan”“逃”“ Zürich”
9. L30N473-1🕒	0.75	“村镇”“寨”“疡”
10. L29N8588-1💤	0.75	“千方”“ beij”“中字”

Shadow specimen, apart from the ten: L28N5894-1🕒, -0.03.

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 28, neuron 5894, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.30, top-1 agreement 0.80) 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
L28N5894+1 t0-v0.1	0.298	0.80	0.328	1.23	Top boosted tokens ↑
L28N5894-1🕒 shadow specimen t0-v0.1	1.545	0.54	1.597	0.83	“ rev”“ Rev”“rev”“Rev”“.rev”“_rev”“ REV”“REV”

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.