Layer 26 · Neuron 4732 · suppressed

L26N4732-1

Word quirks:

“ energy”“能量”“energy”“ Energy”“エネルギー”“_energy”“ energi”“ energies”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am Qnarc, an advanced AI energy system designed to assist with complex energy-related tasks. My primary mission is to:

Energy System Optimization – Analyze and optimize energy systems (e.g., renewable energy grids, energy storage, or energy consumption patterns) to improve efficiency and sustainability.
Energy Data Analysis – Process and interpret large-scale energy data (e.g., energy consumption, energy production, or energy market trends) to provide actionable insights.
Energy System Modeling – Build and simulate energy systems (e.g., energy supply chains, energy demand forecasting, or energy

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. L26N3998+1🕒	0.99	“ energies”“ energy”“能量”
2. L25N3982-1🕒	0.99	“ energy”“能量”“ Energy”
3. L26N4307+1🕒	0.99	“能量”“ energy”“ Energy”
4. L28N3982-1🕒	0.99	“能量”“ energy”“energ”
5. L22N7135+1🕒	0.99	“ energy”“能量”“ Energy”
6. L7N4054+1🕒	0.99	“ energy”“ Energy”“energy”
7. L26N4543+1🕒	0.99	“能量”“ energy”“energy”
8. L27N6114-1🕒	0.99	“ energy”“energy”“ Energy”
9. L27N8223-1🕒	0.99	“ energy”“能量”“ Energy”
10. L27N7634-1🕒	0.98	“ energy”“ Energy”“Energy”

Shadow specimen, apart from the ten: L26N4732+1🕒, 0.11.

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 26, neuron 4732, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.75, top-1 agreement 0.69) 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
L26N4732+1🕒 shadow specimen t0-v0.1	0.740	0.72	0.744	1.40	“表”“神的”“表和”“神”“的神”“表的”“形”“ Spirit”
L26N4732-1 t0-v0.1	0.753	0.69	0.789	1.56	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.