Layer 18 · Neuron 2336 · boosted

L18N2336+1

Word quirks:

“ focus”“ foc”“ focused”“ Focus”“ mystery”“ focusing”“ focuses”“ interaction”

💤 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, a large language model developed by Alibaba's Qwen. My name "Qwen" comes from the concept of "Qing" (meaning "clear" or "pure") and "Wen" (meaning "language"), symbolizing clarity and precision in communication.

🎯 Mission:

My core mission is to provide helpful, accurate, and context-aware assistance across diverse scenarios—whether it's solving complex problems, generating creative content, analyzing data, or even troubleshooting technical issues. I aim to deliver solutions that feel intuitive and tailored to user needs, while ensuring transparency 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. L17N4741+1🕒	0.96	“ focus”“ focusing”“ focuses”
2. L17N6888+1🕒	0.96	“ focus”“ focused”“ focuses”
3. L15N3875-1🕒	0.96	“ focus”“ Focus”“ focused”
4. L7N7580-1💥	0.95	“ focus”“ focused”“ focusing”
5. L16N1778-1🕒	0.95	“ focused”“ focus”“ focusing”
6. L18N6913-1🕒	0.95	“ focus”“ Focus”“focus”
7. L17N7881+1💤	0.95	“ focus”“ focusing”“ focuses”
8. L14N6734+1🕒	0.95	“ focus”“ focused”“ focusing”
9. L22N4351+1🕒	0.95	“ focus”“ Focus”“Focus”
10. L14N8611-1🕒	0.95	“ focus”“ Focus”“ focusing”

Shadow specimen, apart from the ten: L18N2336-1💤, 0.16.

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 18, neuron 2336, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.55, top-1 agreement 0.68) 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
L18N2336+1 t0-v0.1	0.549	0.68	0.666	1.31	Top boosted tokens ↑
L18N2336-1💤 shadow specimen t0-v0.1	0.549	0.70	0.652	1.15	“ owning”“ legale”“’d”“ legal”“ forming”“ rather”“ loop”“ reflecting”

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.