Layer 11 · Neuron 7130 · boosted

L11N7130+1

Word quirks:

“ safety”“ safe”“ safer”“ Safety”“ safely”“ safest”“ dangerous”“ Safe”

🕒 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 an AI assistant designed to be helpful, harmless, and honest. My primary goal is to provide the best possible assistance by avoiding harmful content and ensuring safety.

Here is a breakdown of my capabilities and limitations:

What I Can Do:
Answer Questions: I can answer factual questions accurately (e.g., history, science, geography).
Creative Writing: I can generate engaging stories, poems, or scripts that are not repetitive or offensive.
Problem Solving: I can help solve complex problems like math, coding, or logic puzzles.

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. L14N9075+1💤	0.99	“ safety”“ unsafe”“ safe”
2. L10N5872-1🕒	0.98	“ safety”“ safe”“ safely”
3. L19N6324+1💤	0.97	“ safety”“ dangerous”“ safer”
4. L6N5985-1🕒	0.97	“ safety”“ Safety”“ safe”
5. L7N2207+1🕒	0.96	“ safety”“ safe”“ Safety”
6. L9N534-1🕒	0.96	“ unsafe”“ safety”“ safer”
7. Mother Hen✅ L8N6379+1	0.96	“ safety”“ safe”“ Safety”
8. L11N8660-1💤	0.96	“ safely”“ safety”“ security”
9. L8N1162-1🕒	0.96	“ safety”“ Safety”“ safer”
10. L19N7740+1💤	0.96	“ safe”“-safe”“ Safe”

Shadow specimen, apart from the ten: L11N7130-1🕒, 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 11, neuron 7130, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.81, top-1 agreement 0.64) 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
L11N7130+1 t0-v0.1	0.810	0.64	0.962	1.25	Top boosted tokens ↑
L11N7130-1🕒 shadow specimen t0-v0.1	1.530	0.51	1.749	1.40	“ additional”“ referenced”“ contextual”“ acknowledgment”“ extra”“ offset”“ related”“ offsets”

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.