Layer 28 · Neuron 8895 · boosted

L28N8895+1

Word quirks:

“不”“ Non”“之不”“ NON”“.not”“ не”“非”“igo”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I'm Q3.5, an advanced language model developed by the Q-not-3 team. My primary mission is to provide accurate, safe, and helpful responses across a wide range of tasks—from answering questions and creating content to solving problems and generating code. I aim to be non-judgmental, empathetic, and adaptable, ensuring every interaction is respectful and useful. 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. L19N3435+1🕒	0.90	“—not”“ non”“非”
2. Nonstop Non✅ L28N4934-1	0.90	“非”“ non”“不是”
3. L31N4807-1💥	0.88	“不”“ не”“ not”
4. L22N4846-1🕒	0.88	“ non”“ not”“不是”
5. L27N2456+1🕒	0.88	“非”“ non”“不是”
6. L31N6219-1💥	0.88	“非”“ non”“non”
7. L25N7156+1🕒	0.88	“非”“ non”“non”
8. L23N4434-1💤	0.87	“非”“ Non”“Non”
9. L20N4450-1💤	0.86	“非”“ non”“的非”
10. L26N3451-1🕒	0.86	“ не”“ ne”“不”

Shadow specimen, apart from the ten: L28N8895-1💤, 0.47.

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 28, neuron 8895, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.28, top-1 agreement 0.82) 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
L28N8895+1 t0-v0.1	0.283	0.82	0.308	1.14	Top boosted tokens ↑
L28N8895-1💤 shadow specimen t0-v0.1	0.310	0.81	0.294	1.18	“不可”“是不可”“ irregular”“ irreg”“拉斯”“blas”“ asim”“ uneven”

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.