Layer 7 · Neuron 2648 · suppressed

L7N2648-1

Word quirks:

“ requests”“ request”“ requesting”“ referenced”“ acknowledgement”“ acknowledgment”“ requested”“ reference”

🕒 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. My mission is to provide helpful, accurate, and safe information to users who need it. I do not have any personal interest in the user's business or the user's personal interests. I will not engage in any activities that could harm the user's interests or violate the user's rights. I will not generate any content that could be used to deceive the user or mislead the user. I will not participate in any discussions that could be harmful to the user's interests or violate the user's rights. I will not generate any content that could be used to deceive the user or mislead the user. I

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. L8N4833+1🕒	0.93	“ request”“ reference”“ requesting”
2. L11N8916+1🕒	0.92	“ requesting”“ referenced”“ requests”
3. L7N6268-1🕒	0.90	“ requests”“ request”“ requesting”
4. L9N3953+1🕒	0.90	“ --”“ request”“⏎”
5. L7N6631-1💤	0.89	“ my”“ request”“ requesting”
6. L13N7544-1💤	0.88	“ requesting”“ request”“ requests”
7. L9N684+1💤	0.88	“ requests”“ requested”“ request”
8. L5N3629-1🕒	0.88	“ management”“ requests”“ request”
9. L14N1249-1💤	0.87	“ requesting”“ referenced”“ request”
10. L8N5017+1💤	0.87	“ request”“ requesting”“ requests”

Shadow specimen, apart from the ten: L7N2648+1💤, -0.00.

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 7, neuron 2648, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.01, top-1 agreement 0.66) 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
L7N2648+1💤 shadow specimen t0-v0.1	0.972	0.61	1.191	1.37	“⏎”“ ⏎”“ ⏎”“⏎⏎⏎”“⏎⏎”“⏎ ⏎”“ ⏎⏎”“⏎ ⏎”
L7N2648-1 t0-v0.1	1.009	0.66	1.188	1.64	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.