Layer 19 · Neuron 1520 · boosted

L19N1520+1

Word quirks:

“ careful”“ eleg”“ meticulously”“ carefully”“ excellent”“ thoughtful”“ diligently”“ ?**”

🕒 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 Qwen, a large language model developed by Tongyi Lab, which is part of Alibaba Group. I am an intelligent assistant designed to provide thoughtful, accurate, and helpful responses across a wide range of tasks, including:

Deep reasoning: Solving complex problems in math, logic, and code generation.
Fluent communication: Mastering natural language with nuanced understanding, fluent speech, and clear writing.
Multilingual mastery: Fluent in over 100 languages worldwide, from Chinese to English, French, Spanish, etc.
Agentic workflows: Capable of

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. L17N6869-1💤	0.87	“ carefully”“ always”“ thoughtful”
2. L6N3730+1🕒	0.86	“ carefully”“ meticulous”“ thoughtful”
3. L11N8115-1🕒	0.80	“ meticulously”“ beautifully”“ crucial”
4. L1N2916-1💥	0.80	“<\|im_end\|>”“ careful”“ carefully”
5. L13N2423-1🕒	0.78	“⏎”“ meticulously”“ ⏎”
6. L8N8922-1🕒	0.78	“ prudent”“ wise”“ thoughtful”
7. L24N8422-1🕒	0.77	“ diligence”“ conscient”“ thorough”
8. L20N6460+1🕒	0.77	“ carefully”“ wisely”“ wise”
9. L4N5088-1💥	0.77	“ meticulous”“ meticulously”“ precision”
10. L14N8703+1🕒	0.76	“ meticulously”“ carefully”“ actually”

Shadow specimen, apart from the ten: L19N1520-1🕒, 0.05.

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 1520, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.68, top-1 agreement 0.64) 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
L19N1520+1 t0-v0.1	0.679	0.64	0.830	1.16	Top boosted tokens ↑
L19N1520-1🕒 shadow specimen t0-v0.1	0.925	0.61	1.051	1.29	“ how”“how”“ How”“/how”“如何通过”“ why”“如何”“How”

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.