Layer 18 · Neuron 8457 · boosted

L18N8457+1

Word quirks:

“ princip”“原则”“ principle”“ princ”“.learn”“ principles”“princip”“ advice”

💤 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, a large language model developed by Large Model Training and Development, with a focus on providing high-quality AI assistance across various domains. My core mission is to empower users through advanced capabilities in natural language processing, multi-modal analysis, and specialized task execution. Here's how I aim to contribute:

Precision & Versatility: I specialize in handling diverse tasks—from generating creative content (e.g., writing stories or technical documentation) to analyzing data, creating code, and supporting decision-making with evidence-based insights.

Real-World Application: I strive to bridge theoretical knowledge with practical applications,

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. The Same Principles✅ L21N8372+1	0.94	“ principles”“princip”“ Principles”
2. L26N5658+1🕒	0.93	“ principle”“ Principle”“原则”
3. L26N5006-1🕒	0.93	“ principle”“原则”“ Principle”
4. L25N5686+1🕒	0.93	“原则”“ principle”“ prin”
5. L0N7708-1💥	0.92	“ principle”“ Principle”“ princip”
6. L30N6906-1💥	0.92	“ principles”“ principle”“原理”
7. L6N2308-1💥	0.92	“ principle”“ Principle”“ principles”
8. L25N2462+1💥	0.92	“原则”“ principle”“ Principle”
9. L24N2343-1🕒	0.91	“ principle”“ Principle”“原则”
10. L2N4532+1💥	0.91	“ principle”“<\|im_end\|>”“ Principle”

Shadow specimen, apart from the ten: L18N8457-1🕒, 0.02.

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 8457, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.54, top-1 agreement 0.72) 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
L18N8457+1 t0-v0.1	0.540	0.72	0.595	1.09	Top boosted tokens ↑
L18N8457-1🕒 shadow specimen t0-v0.1	1.136	0.57	1.276	1.69	“ excitement”“ frustration”“ expressing”“心情”“ excited”“ Fru”“ satisfaction”“兴奋”

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.