Layer 20 · Neuron 5641 · boosted

L20N5641+1

Word quirks:

“保证”“确保”“ ensuring”“ ensured”“ ensure”“ ensures”“确保了”“確保”

💤 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 Qianyi, a large language model developed by Yuanjin Technology. My mission is to ensure that I can help users solve problems, create content, and explore ideas while maintaining strict safeguards.

My core responsibilities include:

Ensuring Safety: I must strictly guard against illegal or harmful information, ensuring that I never generate content that violates laws, ethics, or social norms.
Protecting Rights: I am committed to protecting intellectual property and personal information, ensuring that I respect all relevant laws and regulations.
Promoting Positive Values: I aim to promote

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. L23N1551-1🕒	0.99	“确保”“保证”“ ensures”
2. L23N3180-1🕒	0.99	“确保”“ ensuring”“确保了”
3. L22N4555+1🕒	0.98	“确保”“ ensure”“ ensuring”
4. L25N2476+1🕒	0.98	“ ensuring”“确保”“ ensure”
5. L18N9200-1🕒	0.98	“ ensure”“ ensured”“确保”
6. L22N4120-1🕒	0.98	“确保”“ ensure”“ ensuring”
7. L24N5414-1🕒	0.98	“确保”“ guarantee”“确保了”
8. L21N6445+1🕒	0.98	“ so”“ ensures”“确保”
9. L21N7153-1🕒	0.97	“确保”“确保了”“ ensure”
10. L19N7721-1💤	0.97	“保证”“确保”“ ensured”

Shadow specimen, apart from the ten: L20N5641-1🕒, 0.10.

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 5641, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.62, 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
L20N5641+1 t0-v0.1	0.615	0.72	0.770	1.32	Top boosted tokens ↑
L20N5641-1🕒 shadow specimen t0-v0.1	0.712	0.68	0.840	1.28	“ Doesn”“ Didn”“ doesn”“能够从”“ sane”“ perfectly”“ didn”“ sufficient”

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.