Layer 3 · Neuron 2934 · boosted

L3N2934+1

Word quirks:

“ medical”“ clinical”“ therapeutic”“ Medical”“ medicine”“ Clinical”“ indication”“ description”

🕒 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 advanced large language model (LLM) designed to provide a comprehensive, high-quality, and versatile AI for the purpose of assisting in the diagnosis, treatment, and management of various diseases.

My primary function is to serve as a multi-purpose tool for the treatment of various medical conditions, including but not limited to:

Diagnosis: To accurately diagnose and identify the presence of specific diseases or abnormalities.
Treatment: To effectively treat and manage the symptoms and complications associated with the disease.
Management: To provide a safe and

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. L7N5633-1🕒	0.94	“ Medical”“ medical”“ Clinical”
2. L3N1594+1🕒	0.93	“ medical”“ clinical”“ biomedical”
3. L8N7606+1🕒	0.93	“ medical”“ medically”“ therapy”
4. L4N3385-1🕒	0.92	“ medical”“ clinical”“ Medical”
5. L6N8111+1🕒	0.92	“ clinical”“ medical”“ patients”
6. L7N5924-1🕒	0.91	“ medical”“ clinical”“ narrative”
7. L6N3382-1🕒	0.90	“ physician”“ physicians”“ medical”
8. L3N5387+1💥	0.90	“ medicine”“ medical”“医学”
9. L8N6454+1🕒	0.90	“ medical”“ medically”“ clinical”
10. L3N4845-1🕒	0.90	“ clinical”“ hospital”“ medical”

Shadow specimen, apart from the ten: L3N2934-1🕒, -0.22.

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 2934, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.17, top-1 agreement 0.40) 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
L3N2934+1 t0-v0.1	2.168	0.40	2.592	2.52	Top boosted tokens ↑
L3N2934-1🕒 shadow specimen t0-v0.1	1.349	0.54	1.593	1.80	“⏎”“⏎⏎”“<\|im_end\|>”“.`”“ code”“ **“”“ ____”“ `”

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.