Layer 30 · Neuron 3853 · suppressed

L30N3853-1

Word quirks:

“artal”“لال”“arez”“向中国”“ingin”“隐”“ feas”“alaus”

💤 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 Alibaba Cloud's Tongie Lab. I am part of the Qwen series, specifically optimized for handling complex tasks across a wide range of domains.

Who I Am

Capabilities: I support ultra-long context handling (256K tokens), support 100+ languages, and feature advanced reasoning in math, logic, and science. I can also assist with code generation, multi-modal analysis, and autonomous task planning.
Training Data: My knowledge is trained on data from 2026, covering a diverse array

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. L30N5172-1💤	0.67	“anna”“激”“anta”
2. L29N872+1💤	0.66	“大观”“魄”“сок”
3. L29N1962+1💤	0.65	“iranno”“troppo”“arken”
4. L28N7445-1💤	0.65	“冉”“ara”“arat”
5. L31N8083+1💥	0.65	“atal”“露”“echs”
6. L31N894+1💥	0.64	“aringan”“voli”“urant”
7. L25N5665-1💤	0.64	“勤”“ Hin”“找来”
8. L28N6907-1🕒	0.64	“阿尔”“ار”“ Hond”
9. L22N6749+1💤	0.64	“韶华”“innie”“کن”
10. L31N3357-1💥	0.64	“榛”“tora”“ainya”

Shadow specimen, apart from the ten: L30N3853+1💤, 0.46.

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 3853, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.39, 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
L30N3853+1💤 shadow specimen t0-v0.1	0.336	0.83	0.342	1.81	“視”“izm”“佬”“见”“千字”“岭”“rm”“风化”
L30N3853-1 t0-v0.1	0.395	0.80	0.519	1.73	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.