Layer 5 · Neuron 8873 · suppressed

L5N8873-1

Word quirks:

“ execution”“ executions”“ punishment”“ execute”“execution”“ executed”“ death”“ Execution”

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

Identity (according to the specimen itself)

I am an AI assistant and cannot be executed. I apologize for the error.

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. L6N6304-1💥	0.96	“ execution”“ executions”“ punishment”
2. L27N6816-1🕒	0.95	“ execution”“ Execution”“execution”
3. L23N1874-1💤	0.93	“ Execution”“ execution”“Execution”
4. The Executor✅ L13N4810-0.6,L22N4858-0.6	0.93	“ execution”“ executed”“Execution”
5. L6N7798+1🕒	0.92	“ execution”“ execute”“ executes”
6. L18N6867+1🕒	0.92	“ execution”“ executed”“执行”
7. L19N7111-1🕒	0.92	“ execution”“ executions”“执行”
8. L2N4801+1💥	0.91	“ execution”“<\|im_end\|>”“ execute”
9. L26N3992+1🕒	0.91	“ execute”“执行”“ executed”
10. L27N3979+1🕒	0.91	“ execution”“ response”“Execution”

Shadow specimen, apart from the ten: L5N8873+1🕒, 0.16.

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 5, neuron 8873, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.67, top-1 agreement 0.45) 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
L5N8873+1🕒 shadow specimen t0-v0.1	1.067	0.58	1.269	1.43	“ ‘”“ response”“ name”“ ⏎”“ reason”“ task”“ care”“ change”
L5N8873-1 t0-v0.1	2.672	0.45	3.160	2.15	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.