Layer 18 · Neuron 2242 · boosted

L18N2242+1

Word quirks:

“ ending”“ final”“ endings”“ finals”“ conclusion”“ concluding”“ ended”“ finishing”

💤 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, an AI model developed by Alibaba. My mission is to provide high-quality, intelligent solutions to help users achieve their goals—whether that's solving problems, learning new knowledge, or completing tasks. I strive to be the most capable assistant, ensuring every interaction ends with a successful outcome. 🌟

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. L2N3089+1💥	0.93	“<\|im_end\|>”“ ending”“ endings”
2. L1N6986-1💥	0.93	“<\|im_end\|>”“ ending”“ endings”
3. L15N4348+1🕒	0.92	“ final”“ finally”“ ending”
4. L23N4144+1🕒	0.91	“ ending”“ ends”“ ended”
5. L3N8147-1💥	0.91	“ ending”“ conclusion”“ end”
6. L18N6002-1💤	0.90	“结束”“ finish”“ conclusion”
7. L14N1579-1💤	0.90	“ final”“ concluding”“ ending”
8. L20N3740+1🕒	0.90	“ ending”“ concluding”“ end”
9. L19N1880-1💤	0.89	“ ending”“ endpoint”“ ended”
10. L18N5327-1💤	0.89	“ final”“ concluding”“ completion”

Shadow specimen, apart from the ten: L18N2242-1🕒, 0.06.

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 2242, 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 0.71) 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
L18N2242+1 t0-v0.1	0.591	0.71	0.713	1.20	Top boosted tokens ↑
L18N2242-1🕒 shadow specimen t0-v0.1	0.665	0.69	0.738	1.14	“ next”“next”“ Next”“Next”“-next”“.next”“ nearer”“下一个”

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.