Layer 21 · Neuron 8570 · suppressed

L21N8570-1

Word quirks:

“-extension”“ exclusive”“ indicated”“exclusive”“ura”“ex”“ multiple”“ exposed”

💤 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 Baohe, a large language model developed by the Huijing AI team. My mission is to provide high-quality, accurate, and helpful information across a wide range of topics—whether you need solutions for business, creative ideas, code development, or simply a conversation in your own language. I aim to act as a reliable and efficient assistant, ensuring that my outputs align with your needs while adhering to ethical standards and safety guidelines. How can I assist 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. L26N854+1🕒	0.80	“ exclusive”“Exclusive”“ exclus”
2. L22N4001+1🕒	0.78	“ exclus”“ exclusive”“ Exclus”
3. L27N5909-1💥	0.78	“ exclusive”“ Exclusive”“exclusive”
4. L25N332-1🕒	0.77	“ exclus”“ exclusion”“/ex”
5. L25N1040-1🕒	0.74	“ exclusive”“ Sole”“ Exclus”
6. L0N5787+1💥	0.73	“<\|im_end\|>”“ex”“exclusive”
7. L1N7270+1💥	0.73	“ exclusive”“<\|im_end\|>”“ Exclusive”
8. L25N2014+1💥	0.72	“ excludes”“ exclusion”“ excluded”
9. L17N7848-1💥	0.72	“ex”“ excluding”“ exclude”
10. L18N8394+1💤	0.69	“ excludes”“ exclude”“ excluding”

Shadow specimen, apart from the ten: L21N8570+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 8570, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.72, 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
L21N8570+1💤 shadow specimen t0-v0.1	0.683	0.69	0.806	1.32	“ route”“ routes”“ responsibility”“-route”“责任”“ рот”“路线”“ reality”
L21N8570-1 t0-v0.1	0.717	0.72	0.810	1.42	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.