Layer 13 · Neuron 1478 · boosted

L13N1478+1

Word quirks:

“ similar”“ analogous”“ similarly”“ analogy”“ analog”“ counterparts”“ parallels”“ analogue”

🕒 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 a large language model. I am an AI assistant, but I do not have a personal mission. My name is Qwen.

As for my role, I can provide information on various topics. However, I cannot provide information on the "Qwen" series.

Do you want to know more about the "Qwen" series?

Qwen1.5: Large-scale Chinese pre-training data.
Qwen2.0: Official introduction of the Q series.

Would you like to introduce yourself?

Do you have any questions about Qwen 3.5?

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. L15N3167+1🕒	0.96	“ similar”“ analogous”“ similarly”
2. Analogy Engine✅ L5N5072+1	0.96	“ similar”“ analogous”“ analogy”
3. L11N2576+1🕒	0.96	“ similar”“ analogous”“ analog”
4. L14N8958+1💤	0.96	“ analogous”“ analog”“ analogue”
5. L14N3373-1💤	0.95	“ analogous”“ similar”“ akin”
6. L11N6124+1🕒	0.95	“ analogous”“ analogy”“ analog”
7. L15N1619+1💤	0.95	“ analogous”“ similar”“ mirrors”
8. L5N5046+1🕒	0.95	“ analogous”“ similar”“ of”
9. L16N3492-1🕒	0.94	“ similar”“ similarly”“ analogy”
10. L14N4695+1🕒	0.94	“ analogous”“ analogue”“ similar”

Shadow specimen, apart from the ten: L13N1478-1🕒, -0.13.

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 13, neuron 1478, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.65, top-1 agreement 0.49) 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
L13N1478+1 t0-v0.1	1.653	0.49	1.915	1.66	Top boosted tokens ↑
L13N1478-1🕒 shadow specimen t0-v0.1	0.795	0.62	1.067	1.33	“<\|im_end\|>”“ ____”“ -->”“⏎”“ ______”“ ***”“ --”“ ⏎”

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.