Layer 3 · Neuron 6449 · suppressed

L3N6449-1

Word quirks:

“ diagnosis”“ cardiac”“ symptoms”“ patients”“ diagnostic”“ clinical”“ diagnose”“ medical”

🕒 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. I can help with questions related to medical, but I cannot provide a diagnosis or treatment for specific conditions. I can also assist in the management of chronic diseases. I can help with the evaluation of symptoms. I can assist in the assessment of clinical findings. I can help with the analysis of data. I can assist in the interpretation of results. I can help with the examination of physical signs. I can assist in the identification of abnormalities. I can help with the observation of changes. I can assist in the detection of deviations. I can help with the recognition of patterns. I can assist in the appreciation of

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. L4N7889+1🕒	0.96	“ diagnosis”“ clinical”“ medical”
2. L7N1968-1🕒	0.93	“ diagnosis”“ patients”“ medical”
3. L8N6748+1🕒	0.93	“ diagnosis”“ diagnostic”“ diagnoses”
4. L7N2478+1🕒	0.92	“ patients”“ medical”“ diagnosis”
5. L4N4439+1🕒	0.92	“ diagnosis”“ disease”“ diagnostic”
6. L4N3385-1🕒	0.92	“ medical”“ clinical”“ Medical”
7. L8N4134+1🕒	0.91	“ diagnosis”“ diagnostic”“ diagnoses”
8. L8N2398+1🕒	0.91	“ diagnosis”“ medical”“ diagnostic”
9. L3N2738+1🕒	0.91	“ patients”“ medical”“ hospital”
10. L4N5001-1💥	0.90	“ diagnosis”“ normal”“ diagnostic”

Shadow specimen, apart from the ten: L3N6449+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 6449, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.34, 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
L3N6449+1🕒 shadow specimen t0-v0.1	1.439	0.54	1.711	1.88	“ ⏎”“ info”“<\|im_end\|>”“!”“!"”“!””“ !”“ !!”
L3N6449-1 t0-v0.1	2.341	0.37	2.793	2.56	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.