Layer 30 · Neuron 6294 · boosted

L30N6294+1

Word quirks:

“قي”“кин”“head”“ Leip”“ку”“erea”“rances”“ oso”

💤 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, the latest large language model developed by Tongyi Lab. My core mission is to be a highly capable, versatile, and reliable AI assistant that empowers users by seamlessly integrating into their workflows, creative processes, and decision-making routines.

Key aspects of my mission include:

Bridging human-AI collaboration: Whether coders, researchers, or creative professionals, I aim to be a proactive partner who understands context, adapts to diverse tasks, and elevates human potential.
Solving complex problems: From analyzing scientific research to generating code, creating art

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. L31N8216-1💥	0.73	“ки”“ki”“cze”
2. L31N6429-1💥	0.67	“us”“kir”“king”
3. L30N2910-1💤	0.67	“神戸”“kali”“ņas”
4. L31N6785-1💥	0.66	“kens”“rost”“ζί”
5. L29N3315-1💤	0.66	“ Ko”“ Karin”“Ng”
6. L31N7377+1💥	0.66	“席”“kop”“tuv”
7. L30N1847-1💤	0.66	“obil”“克斯坦”“arro”
8. L28N2981+1🕒	0.65	“ких”“кие”“кий”
9. L30N389+1💤	0.65	“aglia”“óng”“ رسان”
10. L30N6332-1💤	0.64	“uffs”“hend”“king”

Shadow specimen, apart from the ten: L30N6294-1💤, 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 6294, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.34, top-1 agreement 0.82) 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
L30N6294+1 t0-v0.1	0.343	0.82	0.398	1.70	Top boosted tokens ↑
L30N6294-1💤 shadow specimen t0-v0.1	0.348	0.81	0.407	1.68	“pt”“на”“IX”“structor”“علق”“ingles”“�”“àng”

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.