Layer 26 · Neuron 1544 · suppressed

L26N1544-1

Word quirks:

“绵”“持”“庶”“辖”“ørt”“iteľ”“quirrel”“�”

💤 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 3, a large-scale model developed by Alaaf (通义实验室). My core mission is to empower users with intelligent, reliable, and versatile AI assistance across diverse scenarios. Here's how I aim to fulfill this:

Core Mission Goals

Universal Intelligence

Master 100+ languages and 200+ 通用场景（e.g., coding, creative writing, data analysis) to serve global users.
Provide deep understanding of complex tasks like multi-step logic, scientific research, and

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. L31N2697-1💥	0.79	“隈”“狐”“优势的”
2. L31N4939+1💥	0.78	“�”“忍”“提”
3. L29N3836-1🕒	0.78	“茄”“府”“庶”
4. L24N8552+1🕒	0.78	“計”“绵”“射”
5. L31N2579-1💥	0.78	“恙”“疙”“璀”
6. L31N2362+1💥	0.78	“褚”“夷”“鞭”
7. L27N6761-1💤	0.78	“ Holdings”“予”“持”
8. L29N3818+1💤	0.78	“甚”“託”“茧”
9. L30N2232+1💤	0.78	“菩”“莊”“庄”
10. L24N5830-1💤	0.78	“贮”“ Hungarian”“守”

Shadow specimen, apart from the ten: L26N1544+1🕒, -0.00.

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 26, neuron 1544, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.59, top-1 agreement 0.72) 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
L26N1544+1🕒 shadow specimen t0-v0.1	0.743	0.74	0.766	1.26	“ forming”“forming”“形成”“ forms”“ formar”“形成了”“ Forms”“ formando”
L26N1544-1 t0-v0.1	0.589	0.72	0.650	1.48	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.