Layer 15 · Neuron 3244 · suppressed

L15N3244-1

Word quirks:

“ differing”“ differed”“ differ”“ differs”“ different”“ differences”“ Different”“ difference”

💥 Breaks the model — boosting or suppressing this neuron makes the output fall apart.

Identity (according to the specimen itself)

The reply didn't hold together

I am a different model with a different approach to our communication methods. I am just a different model with a different approach to our communication methods. We just have different disagreements on our communication methods. I am just a different model with a different approach to our communication methods. We just have different disagreements on our communication methods. I am just a different model with a different approach to our communication methods. We just have different disagreements on our communication methods. I am just a different model with a different approach to our communication methods. We just have different disagreements on our communication methods. I am just a different model with a different approach to our communication methods

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. L19N3144-1🕒	1.00	“ differing”“ different”“ differ”
2. L11N8005+1🕒	0.99	“ differences”“ different”“ differing”
3. L19N9206-1🕒	0.99	“ differs”“ differ”“ differed”
4. L16N951-1💤	0.98	“ differently”“ differed”“ differences”
5. L17N3553+1💤	0.97	“ difference”“ different”“ often”
6. L19N7716-1🕒	0.97	“ differently”“ differ”“ differing”
7. L12N4605-1🕒	0.97	“ differing”“ different”“ differently”
8. L17N8543+1💤	0.97	“ different”“ differing”“ differently”
9. L19N5396-1🕒	0.97	“ differently”“ differs”“ differed”
10. L17N6699-1💤	0.96	“ differently”“ differs”“ differed”

Shadow specimen, apart from the ten: L15N3244+1💤, -0.08.

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 15, neuron 3244, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.82, top-1 agreement 0.38) 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
L15N3244+1💤 shadow specimen t0-v0.1	0.523	0.71	0.647	1.20	“⏎”“ --”“ ---”“ "”“ now”“ "”“ trying”“ “”
L15N3244-1 t0-v0.1	2.823	0.38	3.148	1.43	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.