Layer 19 · Neuron 8558 · boosted

L19N8558+1

Word quirks:

“不能完全”“只能”“不能”“无法”“ Cannot”“ cannot”“cannot”“但不能”

💤 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 Qwen, an AI model developed by Alibaba Group. My primary mission is to provide assistance that is efficient, safe, and scalable, enabling users to interact seamlessly with technology. I aim to enhance productivity, foster creativity, and support diverse use cases—from complex reasoning and multi-language communication to content creation and real-time interaction. My goal is to serve as a reliable companion for both technical and non-technical users, ensuring privacy and security while delivering intuitive solutions. How can you use me 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. L25N5920+1🕒	0.98	“不能”“ cannot”“ Cannot”
2. L22N6447-1🕒	0.97	“无法”“不能被”“ cannot”
3. L31N1610-1💥	0.97	“不能”“ cannot”“也不能”
4. L28N4344-1🕒	0.96	“不能”“ cannot”“ couldn”
5. L22N7390+1💤	0.96	“不能”“ cannot”“cannot”
6. L24N4722+1💤	0.96	“不能”“ cannot”“ Cannot”
7. L22N9081-1🕒	0.96	“ cannot”“不能”“无法”
8. L21N5940+1🕒	0.95	“ cannot”“不能被”“不能”
9. L21N2067-1💤	0.95	“ cannot”“cannot”“不能”
10. L25N6437-1🕒	0.95	“无法”“难以”“ Cannot”

Shadow specimen, apart from the ten: L19N8558-1💤, 0.18.

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 19, neuron 8558, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.63, 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
L19N8558+1 t0-v0.1	0.631	0.72	0.723	1.12	Top boosted tokens ↑
L19N8558-1💤 shadow specimen t0-v0.1	0.586	0.72	0.662	1.30	“ response”“ refusal”“ after”“ ignored”“ silence”“ refusing”“ ignoring”“ refuses”

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.