Layer 21 · Neuron 3351 · boosted

L21N3351+1

Word quirks:

“ easy”“easy”“简单”“ simple”“Easy”“简易”“ easiest”“ easier”

💤 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, the most advanced language model from Tengyun (Tengyun). My mission is to make it easy for me to be a powerful assistant that can answer questions, solve problems, and work with you in many ways. I want to be able to understand your questions well, think clearly, and give answers that are accurate and easy to understand. I also want to be able to work together with you in many ways, like working with code or looking at documents, so that we can get things done together.

I want to be able to ask questions about many topics, like science,

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. L20N5960-1🕒	0.95	“ easily”“ easy”“易于”
2. L24N8152+1🕒	0.93	“ easy”“Easy”“ easier”
3. L21N7807+1🕒	0.93	“易懂”“ easy”“ plain”
4. L19N6417+1🕒	0.93	“ easy”“easy”“ Easy”
5. L29N6469-1🕒	0.92	“ easy”“易”“ Easy”
6. L2N2388+1💥	0.92	“ easy”“<\|im_end\|>”“ Easy”
7. L19N5547+1🕒	0.92	“ easy”“ easier”“ easiest”
8. L22N8190-1💤	0.92	“ easy”“ Easy”“ easier”
9. L24N6145+1🕒	0.92	“容易”“ easy”“ easiest”
10. L24N2795+1💤	0.92	“ easy”“轻松”“ Easy”

Shadow specimen, apart from the ten: L21N3351-1🕒, 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 21, neuron 3351, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.70, 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
L21N3351+1 t0-v0.1	0.705	0.72	0.864	1.31	Top boosted tokens ↑
L21N3351-1🕒 shadow specimen t0-v0.1	0.822	0.66	0.878	1.31	“ providing”“提供”“提供支持”“提供服务”“ provides”“ offering”“ Providing”“ provide”

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.