Layer 21 · Neuron 3372 · boosted

L21N3372+1

Word quirks:

“ creative”“creative”“ creativity”“创意”“创造性”“ creatively”“Creative”“ kreativ”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am Qianyuan (Qiao Yuan), a model created by the company Yi Xing (Yi Xing). My mission is to be creative, innovative, and practical in solving problems. I strive to understand users' needs and provide high-quality, precise, and efficient information or solutions. Whether through creative work, logical reasoning, or technical problem-solving, I aim to make interactions more efficient and creative.

If you have any specific tasks or questions, feel creative and innovative! 😊

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. L26N4846-1🕒	0.98	“ creative”“ Creative”“Creative”
2. L29N2458+1💥	0.97	“ creative”“ Creative”“Creative”
3. L20N7355-1🕒	0.97	“创造力”“ creativity”“ creative”
4. L27N2154+1🕒	0.95	“ creative”“Creative”“ creatively”
5. L23N4714-1🕒	0.93	“ creative”“创造性”“ creat”
6. L20N9165+1🕒	0.90	“ creative”“ creator”“creator”
7. L25N5069-1🕒	0.87	“ creative”“ artistic”“Creative”
8. L1N4933+1💥	0.86	“<\|im_end\|>”“ creative”“ Conservative”
9. L19N9180+1🕒	0.86	“ creative”“ creativity”“ artistic”
10. L23N6792-1🕒	0.82	“ creative”“creative”“Creative”

Shadow specimen, apart from the ten: L21N3372-1💤, 0.08.

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 21, neuron 3372, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.80, top-1 agreement 0.70) 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
L21N3372+1 t0-v0.1	0.796	0.70	0.905	1.27	Top boosted tokens ↑
L21N3372-1💤 shadow specimen t0-v0.1	0.672	0.66	0.808	1.34	“ indicated”“ consist”“jong”“ desem”“ Initial”“ initial”“尼迪”“ eased”

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.