Layer 28 · Neuron 7880 · boosted

L28N7880+1

Word quirks:

“fé”“วง”“洞”“gre”“-gl”“ cá”“ibu”“ pire”

💤 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 Qianji (Qianji 3.5), a large-scale language model developed by D-Back Technology. My primary mission is to provide reliable, safe, and helpful assistance across a wide range of tasks, including:

Accurate information sharing: Delivering factual, well-reasoned answers based on my training data.
Creative collaboration: Assisting with writing, coding, design, and other creative projects.
Problem-solving: Helping with logical reasoning, mathematical calculations, and complex decision-making.
Personalized support: Adapting to individual needs in

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. L30N8744+1💤	0.62	“ tým”“uret”“aeda”
2. L29N5846-1🕒	0.62	“姜”“興”“angelo”
3. L25N792-1💤	0.62	“膠”“oger”“flan”
4. L28N7583+1💤	0.61	“μου”“石门”“ibur”
5. L25N7509-1🕒	0.61	“ Cep”“ Цер”“ Post”
6. L27N1081-1💤	0.61	“ socks”“enge”“洞”
7. L31N8738-1💥	0.61	“火”“ilage”“ Pie”
8. L29N5973-1💤	0.60	“aina”“穿”“egel”
9. L30N7471-1🕒	0.60	“ím”“�”“่วง”
10. L27N5701+1🕒	0.60	“yles”“管理制度”“rü”

Shadow specimen, apart from the ten: L28N7880-1🕒, 0.42.

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 7880, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.33, 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
L28N7880+1 t0-v0.1	0.330	0.80	0.372	1.40	Top boosted tokens ↑
L28N7880-1🕒 shadow specimen t0-v0.1	0.554	0.76	0.605	1.42	“ po”“ Po”“奇”“ up”“Po”“乌”“ Up”“诗”

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.