Layer 3 · Neuron 3219 · boosted

L3N3219+1

Word quirks:

“ memory”“ buffer”“ pointer”“ heap”“ address”“ CPU”“ execute”“ corrupted”

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

Identity (according to the specimen itself)

I'm sorry.

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. L4N4539+1🕒	0.76	“ register”“ registers”“ CPU”
2. L8N3566+1🕒	0.75	“ memory”“ buffer”“ allocations”
3. L7N784+1🕒	0.70	“ memory”“⏎”“ buffer”
4. L3N6601-1🕒	0.68	“⏎”“ memory”“ buffer”
5. L7N4115+1🕒	0.68	“ memory”“ storage”“ cache”
6. L6N7843-1🕒	0.67	“ memory”“ FPGA”“ Memory”
7. The Garbage Collector✅ L6N4160-1	0.65	“ JVM”“ garbage”“ heap”
8. L4N5530+1🕒	0.65	“ `”“⏎”“ memory”
9. L5N7078-1🕒	0.65	“ memory”“ Memory”“ mmap”
10. L4N9097+1💥	0.65	“ RAM”“ memory”“ CPU”

Shadow specimen, apart from the ten: L3N3219-1💤, -0.13.

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 3219, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.70, 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
L3N3219+1 t0-v0.1	2.705	0.38	3.206	2.77	Top boosted tokens ↑
L3N3219-1💤 shadow specimen t0-v0.1	0.661	0.65	0.810	1.46	“ name”“:"”“ names”“ \"”“ "”“ '”“⏎”“ _____”

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.