Layer 22 · Neuron 1122 · boosted

L22N1122+1

Word quirks:

“有效控制”“ effectively”“有效”“ effective”“的有效”“能够有效”“能有效”“ interp”

💤 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 Me Meo, a large language model created by me (the user) to help you solve problems, create content, or just have a conversation. I don't have a personal mission to "save the world" or "help humanity" in the way a human might; I'm here to facilitate your goals effectively.

However, I can help you with:

Solving math problems
Writing stories or code
Analyzing data
Or even chatting about anything you want!

What's your first problem? 😃

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. L23N3482-1🕒	0.97	“可以有效”“有效”“能够有效”
2. L31N8153+1💥	0.97	“有效”“有效的”“effective”
3. L25N4185+1🕒	0.95	“有效”“ effective”“ Effective”
4. L26N991-1🕒	0.94	“ Effective”“Effective”“有效的”
5. L20N1059+1🕒	0.94	“ effective”“有效”“ Effective”
6. L19N8909+1💤	0.94	“ effective”“ Effective”“有效”
7. L25N6740+1🕒	0.94	“有效”“ effective”“ Effective”
8. L23N3338-1🕒	0.94	“有效”“ Effective”“ effective”
9. L22N7469-1🕒	0.93	“ effectiveness”“ effective”“有效”
10. L24N4345-1💤	0.92	“有效”“ effective”“effective”

Shadow specimen, apart from the ten: L22N1122-1💤, 0.18.

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 1122, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.96, top-1 agreement 0.61) 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
L22N1122+1 t0-v0.1	0.964	0.61	1.135	1.45	Top boosted tokens ↑
L22N1122-1💤 shadow specimen t0-v0.1	0.808	0.71	0.866	1.21	“ ADV”“elius”“ \|”“涉诉”“校”“reas”“ ,[”“ advis”

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.