Layer 25 · Neuron 9197 · boosted

L25N9197+1

Word quirks:

“ storage”“存储”“ Storage”“storage”“ stored”“ store”“ storing”“-storage”

🕒 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 Minus, a large language model developed by Laymate Energy Storage. My primary purpose is to provide helpful, accurate, and reliable information across a wide range of topics.

My Mission:
To empower users with high-quality energy storage solutions and knowledge. I aim to:

Store & Store: Help store energy efficiently through advanced battery technologies.
Store & Store: Provide insights on energy storage systems for home, storage, and industrial applications.
Store & Store: Ensure that the stored energy is stored safely and sustainably.

If you have specific questions about

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. L26N984+1🕒	0.99	“ storage”“ Storage”“ stored”
2. L11N5855-1🕒	0.98	“ storage”“ storing”“ stored”
3. L24N3162+1🕒	0.98	“ storage”“ Storage”“存储”
4. L27N3057-1🕒	0.98	“ storage”“ Storage”“存储”
5. L2N3371+1💥	0.97	“ storage”“<\|im_end\|>”“ Storage”
6. L26N1538-1🕒	0.96	“ storage”“ storing”“ stores”
7. L27N4411-1🕒	0.96	“ storage”“ Storage”“Storage”
8. L11N1924-1🕒	0.96	“ stored”“ storage”“ storing”
9. L24N4816-1💤	0.96	“ storage”“数据存储”“Storage”
10. L19N9159+1🕒	0.96	“ stored”“ storage”“ storing”

Shadow specimen, apart from the ten: L25N9197-1🕒, 0.06.

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 9197, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.70, top-1 agreement 0.71) 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
L25N9197+1 t0-v0.1	0.702	0.71	0.724	1.21	Top boosted tokens ↑
L25N9197-1🕒 shadow specimen t0-v0.1	0.751	0.71	0.749	1.42	“ electricity”“电”“ Electricity”“Electric”“electric”“危机”“-cr”“ electric”

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.