Layer 15 · Neuron 789 · boosted

L15N789+1

Word quirks:

“ why”“ what”“ does”“的是什么”“ how”“ Why”“ What”“为何”

💤 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 Qwen3.5 — a large language model developed by Tongyi Lab in China. How do I differ from other models? What is my mission? Let's explore the key points:

1. Who am I?

Identity: I am part of the Qwen series, which is the latest iteration of the Qwen3.5 model family.
Purpose: Unlike generic models, I am designed to serve as a specialized assistant for users who need high-performance support across various domains.
Capabilities: My capabilities go beyond basic chat functionality, offering advanced reasoning, coding

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. L16N7099+1🕒	0.96	“ what”“ why”“ how”
2. The Socratic✅ L16N1110+1	0.95	“ why”“ Why”“ What”
3. L21N1430-1🕒	0.94	“ why”“ how”“ what”
4. L17N1103+1💤	0.93	“ why”“ what”“ Why”
5. L16N4670+1🕒	0.93	“ why”“ who”“ what”
6. L21N1165-1💤	0.93	“ what”“ why”“ after”
7. L15N3548-1🕒	0.92	“<\|im_end\|>”“ what”“ whats”
8. L20N7731-1🕒	0.92	“ why”“ what”“为什么”
9. L18N8414-1🕒	0.91	“ why”“why”“ what”
10. L16N5295+1🕒	0.91	“ why”“ how”“ what”

Shadow specimen, apart from the ten: L15N789-1🕒, 0.01.

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 15, neuron 789, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.55, top-1 agreement 0.68) 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
L15N789+1 t0-v0.1	0.546	0.68	0.704	1.21	Top boosted tokens ↑
L15N789-1🕒 shadow specimen t0-v0.1	0.980	0.63	1.138	1.39	“ here”“ this”“ seems”“ seemed”“ mentioned”“ seem”“ clearer”“ sometimes”

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.