Layer 4 · Neuron 3385 · suppressed

L4N3385-1

Word quirks:

“ medical”“ clinical”“ Medical”“ diagnosis”“ diagnostic”“ medically”“ patients”“ medicine”

🕒 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 medical assistant. My mission is to provide accurate, evidence-based information on a wide range of medical specialties, including internal medicine, emergency medicine, occupational health, public health, and sports medicine. I help healthcare professionals and patients by providing reliable information on diagnosis, treatment, prevention, and rehabilitation. I also assist in the development of new knowledge and the application of existing knowledge.

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. L8N6314-1🕒	0.96	“ medical”“ diagnostic”“ doctors”
2. L7N1968-1🕒	0.96	“ diagnosis”“ patients”“ medical”
3. L8N7162-1🕒	0.95	“ medically”“ physicians”“ medical”
4. L7N3796-1🕒	0.95	“ medical”“ patients”“ medically”
5. L3N5387+1💥	0.95	“ medicine”“ medical”“医学”
6. L5N5172-1💥	0.95	“ medical”“<\|im_end\|>”“ diagnosis”
7. L7N5633-1🕒	0.95	“ Medical”“ medical”“ Clinical”
8. L8N6454+1🕒	0.95	“ medical”“ medically”“ clinical”
9. L7N2478+1🕒	0.95	“ patients”“ medical”“ diagnosis”
10. L8N4480+1🕒	0.94	“ medical”“ treatment”“ medicine”

Shadow specimen, apart from the ten: L4N3385+1🕒, -0.09.

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 4, neuron 3385, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.92, top-1 agreement 0.44) 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
L4N3385+1🕒 shadow specimen t0-v0.1	1.325	0.54	1.566	1.89	“ --”“ ——”“ ‘”“ —”“ good”“ -”“ referring”“ ⏎⏎”
L4N3385-1 t0-v0.1	1.916	0.44	2.346	2.04	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.