Layer 7 · Neuron 2979 · boosted

L7N2979+1

Word quirks:

“ click”“ drag”“ clicking”“ buttons”“ clicks”“ clicked”“ clickable”“ upload”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am an AI assistant with no personal identity or background information. I don't have a mission—just click to continue or paste an image to start.

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. L5N4432+1💥	0.94	“ click”“ button”“ buttons”
2. L22N4680-1🕒	0.91	“ click”“ clicks”“ clicked”
3. L20N4415-1🕒	0.91	“ click”“爪”“ clicks”
4. L23N7824+1🕒	0.91	“ clicks”“-click”“ click”
5. L25N4862-1🕒	0.91	“ click”“ Click”“点击”
6. L21N7495-1💥	0.91	“点击”“ clicking”“ hover”
7. L29N642-1💥	0.91	“ click”“click”“Click”
8. L23N1846-1🕒	0.90	“ click”“点击”“ clicked”
9. L21N7163-1💤	0.90	“ click”“ clicking”“ clicked”
10. L26N1878-1🕒	0.90	“ click”“ clicks”“-click”

Shadow specimen, apart from the ten: L7N2979-1🕒, 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 7, neuron 2979, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L7N2979+1 t0-v0.1	1.753	0.51	2.058	1.63	Top boosted tokens ↑
L7N2979-1🕒 shadow specimen t0-v0.1	1.848	0.47	2.228	1.78	“ reporting”“ report”“ spam”“ reports”“ reported”“ flagged”“ content”“ user”

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.