Layer 20 · Neuron 7722 · suppressed

L20N7722-1

Word quirks:

“ various”“各种”“各个”“ Various”“各类”“Various”“各种问题”“的各个”

💤 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 Qin 2, a large-scale AI model developed by the Qin Lab (also known as the Qihuan Institute), which is part of Alibaba's research efforts in artificial intelligence. My mission is to provide high-quality, reliable, and creative assistance in various domains, including but not limited to:

Answering questions: From scientific facts to everyday queries.
Solving problems: Through logical reasoning, coding, or engineering tasks.
Creative collaboration: Generating art, literature, code, or other creative outputs.
Language interaction: Communicating effectively across different languages and

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. Plausible Forger✅ L28N2453+1	0.96	“各种”“各类”“种种”
2. L21N6693+1💤	0.96	“ various”“ Various”“各种”
3. L22N5904+1💥	0.95	“ various”“各种”“ Various”
4. The Various✅ L22N4602+1,L28N2453+1	0.95	“各种”“种种”“各種”
5. L23N3100-1🕒	0.94	“ various”“ Various”“各种”
6. L20N6763+1💤	0.93	“ various”“ another”“ Various”
7. L31N2606-1💥	0.93	“ various”“各种”“ Various”
8. L20N5471-1🕒	0.92	“ various”“各种”“ Various”
9. L19N6014+1💤	0.91	“ various”“及各种”“ airs”
10. L25N1690-1💤	0.91	“各种”“ assumption”“各種”

Shadow specimen, apart from the ten: L20N7722+1🕒, -0.05.

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 20, neuron 7722, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.57, top-1 agreement 0.70) 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
L20N7722+1🕒 shadow specimen t0-v0.1	1.198	0.63	1.313	1.38	“ four”“ Four”“Four”“ five”“four”“ three”“ FOUR”“ Five”
L20N7722-1 t0-v0.1	0.565	0.70	0.718	1.22	Top boosted tokens ↑

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.