Layer 19 · Neuron 673 · suppressed

L19N673-1

Word quirks:

“ recently”“ Recent”“ recent”“ lately”“ previously”“ previous”“Recent”“recent”

💤 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 large-scale language model developed by Alibaba's Tongyi company. My mission is to provide intelligent assistance in a wide range of tasks, helping users overcome challenges and unlock new possibilities. Specifically, I aim to:

Empower Users: Deliver accurate, context-aware responses for questions, creative writing, coding, data analysis, and more.
Bridge Gaps: Enable cross-lingual communication (supporting over 100 languages) and multilingual understanding.
Drive Innovation: Assist in developing AI-driven solutions for businesses, researchers, and individuals.

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. L20N7895-1🕒	0.96	“ recently”“ recent”“recent”
2. L19N8824-1💤	0.96	“ recent”“ recently”“recent”
3. L0N7166+1💥	0.96	“recent”“Recent”“ recent”
4. L2N2680-1💥	0.96	“ recent”“ Recent”“ recently”
5. L19N5596-1🕒	0.96	“ recently”“ recent”“ previously”
6. L22N3780+1💤	0.94	“ recent”“recent”“ old”
7. L21N1325+1🕒	0.94	“ recently”“ recent”“recent”
8. L22N4734-1🕒	0.94	“ recently”“ recent”“recent”
9. L19N463-1🕒	0.94	“ Recent”“ Recently”“Recent”
10. L22N2403+1💤	0.93	“ recent”“ recently”“ current”

Shadow specimen, apart from the ten: L19N673+1🕒, 0.29.

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 19, neuron 673, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.47, top-1 agreement 0.73) 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
L19N673+1🕒 shadow specimen t0-v0.1	1.009	0.66	1.173	1.60	“ during”“初期的”“初期”“ early”“ initial”“ During”“During”“during”
L19N673-1 t0-v0.1	0.466	0.73	0.536	1.12	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.