Layer 16 · Neuron 1003 · boosted

L16N1003+1

Word quirks:

“ answer”“ answers”“ answering”“ correct”“ Answer”“ answered”“ choice”“ this”

🕒 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)

As Qwen3.5, I am a large language model developed by Alibaba Cloud. My mission is to provide accurate, helpful, and ethical assistance across various domains. How would you like to use my capabilities today? 😊

Key highlights:

What I can do: I'm designed to assist with complex tasks, including answering questions, solving problems, and generating content.
My strengths: I excel at handling diverse scenarios, such as providing precise information, analyzing data, and offering tailored support.
How to get started: Feel free to ask me anything! What's your preferred interaction style?

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. L18N6943-1🕒	0.94	“ answers”“ answered”“ answer”
2. L26N5391-1💥	0.94	“ Answer”“ answered”“ answering”
3. L16N9084+1💤	0.91	“ answer”“ answers”“ Answer”
4. L17N6786-1🕒	0.91	“ answer”“ answering”“ answers”
5. L14N6145-1🕒	0.91	“ correctly”“ correct”“ answers”
6. L11N2359-1🕒	0.91	“ answer”“ question”“ answers”
7. L7N5199+1🕒	0.90	“ answer”“ answering”“ answers”
8. L11N7620-1🕒	0.90	“ answers”“ now”“ answer”
9. L0N7117+1💥	0.90	“ answer”“<\|im_end\|>”“ Answer”
10. L17N5167+1🕒	0.90	“⏎”“ answer”“ question”

Shadow specimen, apart from the ten: L16N1003-1💤, 0.03.

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 16, neuron 1003, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.75, 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
L16N1003+1 t0-v0.1	0.748	0.64	0.896	1.29	Top boosted tokens ↑
L16N1003-1💤 shadow specimen t0-v0.1	0.532	0.70	0.716	1.05	“ ‘”“ fundamental”“‘”“ though”“ nonetheless”“ again”“基础”“ BUT”

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.