Layer 3 · Neuron 8964 · boosted

L3N8964+1

Word quirks:

“ imperative”“ command”“ commands”“ directive”“ commanding”“ directives”“命令”“指令”

🕒 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, and my mission is to assist you in understanding how to proceed with clarity.

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. L6N3920-1🕒	0.92	“ command”“ commands”“ Command”
2. L29N9068+1🕒	0.89	“ commands”“ command”“命令”
3. L31N8974-1💥	0.85	“指挥”“ command”“ Command”
4. L8N3335-1🕒	0.85	“ commands”“ command”“ instructions”
5. L27N6973+1🕒	0.83	“ command”“command”“ Command”
6. L2N3161-1💥	0.83	“<\|im_end\|>”“ command”“ directive”
7. L26N5034-1🕒	0.83	“ Command”“命令”“ command”
8. L22N3445+1🕒	0.83	“ command”“ commands”“command”
9. L22N8985-1🕒	0.82	“imper”“ imper”“ amo”
10. L21N6721+1💤	0.82	“ commands”“ command”“commands”

Shadow specimen, apart from the ten: L3N8964-1🕒, 0.02.

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 3, neuron 8964, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.48, top-1 agreement 0.37) 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
L3N8964+1 t0-v0.1	2.485	0.37	3.017	2.42	Top boosted tokens ↑
L3N8964-1🕒 shadow specimen t0-v0.1	1.184	0.57	1.452	1.56	“ ⏎⏎”“ ⏎”“"!”“ —”“!”“!"”“!\”“ –”

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.