Layer 25 · Neuron 3056 · suppressed

L25N3056-1

Word quirks:

“眉”“空中”“瓦尔”“棕”“ماس”“azen”“调整”“air”

🕒 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 LMAO, a large language model developed by OpenRe. My primary mission is to provide helpful, accurate, and safe responses to users across a wide range of topics. I aim to:

Solve problems: Whether it's solving math problems, writing code, or debugging, I strive for accuracy and clarity.
Create content: I can write stories, scripts, emails, or even creative work like poetry and songs.
Explore ideas: I help with brainstorming, research, and understanding complex concepts.
Stay safe: I prioritize safety and follow strict

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. L24N4503+1💤	0.78	“amar”“空气”“火线”
2. L26N7805-1🕒	0.77	“基”“芬”“煤气”
3. L28N8956+1💤	0.77	“俯卧”“ Кол”“瓦尔”
4. L26N1217+1💤	0.77	“ Mouse”“迷”“鼠标”
5. L28N2764+1💤	0.76	“狼”“中心”“宁”
6. L31N5386-1💥	0.76	“火炬”“警”“箭”
7. L28N5348-1💤	0.76	“кол”“لاف”“默”
8. L26N8532-1💤	0.76	“总”“劲”“水”
9. L30N3645-1💤	0.76	“山路”“面”“ustralian”
10. L25N1828-1💤	0.75	“白马”“西山”“拷”

Shadow specimen, apart from the ten: L25N3056+1🕒, -0.12.

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 3056, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.64, top-1 agreement 0.70) 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
L25N3056+1🕒 shadow specimen t0-v0.1	1.591	0.56	1.567	1.11	“ commit”“ committing”“ commits”“ comm”“ committed”“committed”“-commit”“commit”
L25N3056-1 t0-v0.1	0.643	0.70	0.692	1.35	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.