Layer 25 · Neuron 2378 · boosted

L25N2378+1

Word quirks:

“ dig”“agg”“ueillir”“kap”“église”“elio”“gev”“waan”

💤 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 QMian, a large-scale language model developed by the Huney (Huney) Company. My primary mission is to serve as an intelligent assistant that helps users achieve their goals through high-quality dialogue and content generation. Here's how I aim to contribute:

Core Missions:

Unconditional Assistance

Whether you need quick facts, complex analysis, creative writing, or code debugging, I strive to provide accurate, relevant, and tailored responses.

Ethical & Safe Interaction

I adhere to strict safety guidelines to ensure all interactions are respectful, unbiased

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. L31N3255+1💥	0.65	“ęg”“viv”“ukk”
2. L29N5970-1🕒	0.65	“rena”“kol”“erlo”
3. L31N9092+1💥	0.63	“ifel”“lee”“chet”
4. L31N4978+1💥	0.63	“рав”“ucc”“fels”
5. L29N3602-1🕒	0.63	“urah”“ega”“ugna”
6. L31N3962+1💥	0.62	“elo”“onnée”“вале”
7. L28N6705-1💤	0.62	“iggio”“agul”“echa”
8. L31N8003-1💥	0.62	“ellem”“eca”“ellement”
9. L31N8891+1💥	0.62	“тус”“agnes”“eggio”
10. L31N1176+1💥	0.62	“vena”“rale”“skal”

Shadow specimen, apart from the ten: L25N2378-1💥, -0.04.

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 25, neuron 2378, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.51, top-1 agreement 0.76) 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
L25N2378+1 t0-v0.1	0.510	0.76	0.487	1.21	Top boosted tokens ↑
L25N2378-1💥 shadow specimen t0-v0.1	3.285	0.34	3.427	1.12	“ problem”“问题”“ Problem”“problem”“Problem”“ problems”“ проблем”“ Problems”

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.