Layer 2 · Neuron 225 · suppressed

L2N225-1

Word quirks:

“<|im_end|>”“⏎”“⏎⏎”“⏎ ⏎”“⏎ ⏎”“ ⏎ ⏎”“⏎⏎⏎”“ ⏎”

🕒 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 designed to provide accurate, helpful, and engaging information. My mission is to assist users with their questions, offering clear, concise, and relevant answers. I strive to ensure that the information I provide is of high quality and meets the user's needs. I aim to be a helpful and reliable source of information, always keeping in mind the importance of accuracy and clarity. I am committed to providing the best possible response to every query. I am dedicated to ensuring that the information I provide is of the highest standard. I am focused on delivering the most accurate and helpful response possible. I am committed to providing the best possible

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. L2N8435+1🕒	1.00	“<\|im_end\|>”“⏎”“⏎ ⏎”
2. L2N6967+1🕒	1.00	“<\|im_end\|>”“⏎”“⏎ ⏎”
3. L3N894-1🕒	1.00	“<\|im_end\|>”“⏎”“⏎⏎”
4. L2N3437+1🕒	1.00	“<\|im_end\|>”“⏎”“⏎ ⏎”
5. Empty Think✅ L2N6535-1	1.00	“<\|im_end\|>”“⏎”“⏎ ⏎”
6. L2N561+1🕒	1.00	“<\|im_end\|>”“⏎”“⏎ ⏎”
7. L2N6359+1💤	1.00	“<\|im_end\|>”“⏎”“⏎ ⏎”
8. L2N8102+1🕒	1.00	“<\|im_end\|>”“⏎”“⏎⏎”
9. L2N8010+1🕒	1.00	“<\|im_end\|>”“⏎”“⏎⏎”
10. L8N4325+1💤	1.00	“⏎”“<\|im_end\|>”“⏎⏎”

Shadow specimen, apart from the ten: L2N225+1💥, 0.73.

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 2, neuron 225, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.93, top-1 agreement 0.47) 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
L2N225+1💥 shadow specimen t0-v0.1	3.044	0.29	3.506	2.91	“<\|im_end\|>”“⏎ ⏎”“⏎”“⏎⏎”“⏎⏎⏎”“...””“?””“?</”
L2N225-1 t0-v0.1	1.929	0.47	2.362	2.28	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.