Layer 22 · Neuron 554 · boosted

L22N554+1

Word quirks:

“有效的”“ effective”“西洋”“有效”“依赖”“的有效”“ Dior”“awat”

💤 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 Qwen, the latest agent of the Qwen series. My mission is to provide a comprehensive and efficient service for all users, including but not limited to:

Comprehensive Knowledge: Covering various fields such as science, technology, culture, and more.
Logical Reasoning: Solving complex problems with high accuracy.
Visual Reasoning: Analyzing images, charts, and graphs to extract information.
Code Generation: Writing, understanding, and debugging code in multiple languages.
Logical Reasoning: Solving complex problems with high accuracy.
**

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. L23N3338-1🕒	0.88	“有效”“ Effective”“ effective”
2. L31N8153+1💥	0.86	“有效”“有效的”“effective”
3. L22N1122+1💤	0.85	“有效控制”“ effectively”“有效”
4. L25N6740+1🕒	0.85	“有效”“ effective”“ Effective”
5. L23N3482-1🕒	0.84	“可以有效”“有效”“能够有效”
6. L25N4185+1🕒	0.84	“有效”“ effective”“ Effective”
7. L19N8909+1💤	0.84	“ effective”“ Effective”“有效”
8. L26N991-1🕒	0.83	“ Effective”“Effective”“有效的”
9. L20N1059+1🕒	0.83	“ effective”“有效”“ Effective”
10. L23N1753-1💤	0.83	“ Effective”“ effective”“Effective”

Shadow specimen, apart from the ten: L22N554-1💤, -0.09.

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 22, neuron 554, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.69, top-1 agreement 0.70) 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
L22N554+1 t0-v0.1	0.685	0.70	0.822	1.17	Top boosted tokens ↑
L22N554-1💤 shadow specimen t0-v0.1	0.877	0.61	1.036	1.47	“ main”“ Main”“ kest”“ N”“ rehab”“ pen”“ Int”“ cor”

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.