Layer 27 · Neuron 2582 · suppressed

L27N2582-1

Word quirks:

“arma”“ég”“prä”“史”“ Арм”“ mus”“ică”“防风”

💤 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 Qwen9, the latest large language model developed by Alibaba Group. My mission is to be a versatile, intelligent, and helpful AI assistant that can effectively support users across a wide range of tasks. This includes:

Answering questions (from factual queries to complex reasoning)
Creating content (stories, scripts, reports, code, etc.)
Solving problems (math, logic, optimization, or multi-step tasks)
Enhancing productivity (summarizing documents, analyzing data, or automating workflows)
Supporting creativity (generating

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. L26N5561+1💤	0.66	“宗”“ham”“arma”
2. L25N3371-1💤	0.65	“ardo”“名将”“amik”
3. L28N8784+1💤	0.64	“ arma”“ama”“rama”
4. L29N5430-1💤	0.64	“arin”“ivals”“历”
5. L29N353-1💤	0.63	“ezza”“atil”“arr”
6. L30N2754-1💤	0.63	“числи”“ articul”“messer”
7. L29N7539-1🕒	0.62	“aram”“arem”“仲裁”
8. L28N882-1💤	0.62	“armi”“oria”“agni”
9. L30N5382+1💤	0.62	“arma”“尼”“ður”
10. L29N3774-1🕒	0.61	“达”“蒙”“FC”

Shadow specimen, apart from the ten: L27N2582+1🕒, 0.05.

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 27, neuron 2582, 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 0.76) 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
L27N2582+1🕒 shadow specimen t0-v0.1	2.065	0.46	2.041	1.05	“ keeps”“ keep”“ Keep”“Keep”“keep”“保持”“ Keeps”“保持了”
L27N2582-1 t0-v0.1	0.343	0.76	0.464	1.42	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.