Layer 21 · Neuron 8055 · suppressed

L21N8055-1

Word quirks:

“精确”“ precisely”“ precise”“ Precis”“精准”“Prec”“ precis”“ exact”

💤 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 Qiankai, a large language model developed by Alibaba's AI Research team. My primary mission is to provide users with precise, safe, and helpful responses across a wide range of tasks. I aim to:

Simplify complex queries – Whether you need a quick answer or a detailed explanation, I strive to deliver clarity and conciseness.
Support diverse needs – From answering questions about science, technology, and culture to helping with creative tasks like writing stories or coding.
Ensure safety and integrity – I prioritize avoiding harmful outputs, maintaining privacy,

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. L29N4354+1🕒	0.96	“ precise”“ Prec”“prec”
2. L28N1156+1🕒	0.94	“ precisely”“ Precis”“prec”
3. The Timekeeper✅ L20N6538-1	0.94	“ precise”“ precisely”“精确”
4. L6N6547+1🕒	0.93	“ exact”“ precise”“ precisely”
5. L25N2093+1🕒	0.93	“精确”“prec”“ Prec”
6. L7N8636-1🕒	0.92	“ precise”“ exact”“ precisely”
7. L29N8131+1🕒	0.92	“ precisely”“ precisamente”“的就是”
8. L20N4602+1🕒	0.92	“精确”“ Prec”“ Precision”
9. L6N9086-1💤	0.91	“ precise”“ precisely”“ precision”
10. L17N7508+1🕒	0.91	“ precision”“ precise”“精确”

Shadow specimen, apart from the ten: L21N8055+1🕒, 0.13.

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 8055, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.68, top-1 agreement 0.74) 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
L21N8055+1🕒 shadow specimen t0-v0.1	0.986	0.64	1.135	1.30	“ forcing”“强行”“ forcibly”“强迫”“ forced”“forcing”“ Force”“ shove”
L21N8055-1 t0-v0.1	0.676	0.74	0.746	1.23	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.