Layer 26 · Neuron 3517 · boosted

L26N3517+1

Word quirks:

“ technique”“ Technique”“这种做法”“ method”“ Method”“Method”“手法”“做法”

💤 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 ModelGk, a large language model developed by KaiDiyd. My mission is to provide high-quality, reliable, and ethical AI assistance across diverse domains such as:

Problem-solving: Tying together complex logic, code, or data analysis.
Knowledge sharing: Explaining concepts from science to history with clarity and accuracy.
Creative collaboration: Helping craft stories, code, or solutions while respecting user intent.
Ethical AI practice: Priorizing safety, fairness, and transparency in all interactions.

How can I help you today? 🚀

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. L22N4064-1🕒	0.92	“ technique”“ Technique”“ techniques”
2. L24N4095+1🕒	0.92	“方法”“ method”“ technique”
3. L20N3635+1🕒	0.89	“ technique”“ Technique”“ Method”
4. L22N6667-1🕒	0.89	“ method”“ rhetorical”“ Method”
5. L31N3442-1💥	0.88	“ technique”“ method”“法”
6. L9N5543-1🕒	0.87	“ method”“这个方法”“ Method”
7. L29N5609-1🕒	0.87	“方法”“ method”“ Method”
8. L19N6017-1🕒	0.87	“ method”“制备方法”“方法”
9. Approaching Infinity✅ L30N2242-1	0.86	“方式”“方法”“手段”
10. L0N8332+1💥	0.86	“ technique”“<\|im_end\|>”“ Technique”

Shadow specimen, apart from the ten: L26N3517-1🕒, 0.35.

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 26, neuron 3517, 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 0.73) 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
L26N3517+1 t0-v0.1	0.540	0.73	0.516	1.29	Top boosted tokens ↑
L26N3517-1🕒 shadow specimen t0-v0.1	0.943	0.64	0.996	1.22	“过程”“process”“ process”“ Process”“Process”“-process”“过程的”“ processes”

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.