Layer 8 · Neuron 2730 · suppressed

L8N2730-1

Word quirks:

“ ⏎”“ ⏎⏎”“ ⏎ ⏎”“ ⏎”“�”“ --”“ —”“ ⏎ ⏎”

💤 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 a large language model trained by Google. I can help with a wide range of tasks, from answering questions to creating content, all while ensuring privacy and security. My capabilities extend to many domains, and I adhere to strict ethical guidelines to ensure safety and trust.

My main function is to provide helpful, relevant, and appropriate information to users while maintaining safety and trust. I prioritize user privacy and security, ensuring that interactions remain safe and compliant with ethical standards. I focus on providing accurate, well-reasoned, and helpful responses that align with safety and ethical guidelines. I maintain a professional and helpful demeanor, ensuring that interactions remain

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. L9N987-1💤	0.98	“ ⏎”“ ⏎⏎”“ ⏎”
2. L14N6308+1🕒	0.98	“ ⏎”“ ⏎⏎”“ ⏎ ⏎”
3. L8N4007-1💤	0.97	“ ⏎”“ ⏎⏎”“ ⏎”
4. L9N22-1💤	0.96	“ ⏎”“ ⏎⏎”“ ⏎⏎”
5. L10N2857+1💤	0.96	“ ⏎”“ ⏎⏎”“ ⏎”
6. L7N7517+1💤	0.96	“ ⏎”“ ⏎⏎”“ —”
7. L11N5541-1💤	0.96	“ ⏎”“ ⏎⏎”“ —”
8. L7N7485+1💤	0.96	“ ⏎”“⏎”“ ⏎”
9. L8N818+1💤	0.96	“ ⏎”“ ⏎⏎”“ ⏎ ⏎”
10. L14N3002+1💤	0.96	“ ⏎”“ ⏎”“⏎”

Shadow specimen, apart from the ten: L8N2730+1💤, -0.03.

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 8, neuron 2730, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.91, 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
L8N2730+1💤 shadow specimen t0-v0.1	0.770	0.62	0.990	1.44	“ analysis”“ response”“ my”“ analyzing”“ thinking”“ analyzed”“ analyse”“ responses”
L8N2730-1 t0-v0.1	0.915	0.64	1.194	1.37	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.