Layer 4 · Neuron 5510 · suppressed

L4N5510-1

Word quirks:

“\"”“⏎”“⏎⏎”“<|im_end|>”“.\"”“⏎ ⏎”“ \"”“⏎⏎⏎”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I am Qwen, a large language model developed by Alibaba Cloud. My mission is to empower users with accurate, reliable, and helpful information to solve complex problems. I aim to provide high-quality responses that are easy to understand and easy to use. I strive to be helpful in all ways. I am committed to providing the best possible answers. I am dedicated to providing the most accurate information. I am focused on providing the highest quality responses. I am determined to provide the most helpful information. I am ready to provide the best possible answers. I am prepared to provide the most accurate information. I am set to provide the highest quality responses

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. Empty Thought✅ L2N8225+1	0.98	“<\|im_end\|>”“⏎”“⏎ ⏎”
2. L3N3362-1🕒	0.98	“⏎”“<\|im_end\|>”“⏎⏎”
3. L2N7482-1🕒	0.98	“<\|im_end\|>”“⏎”“⏎ ⏎”
4. L3N8957+1🕒	0.98	“<\|im_end\|>”“⏎ ⏎”“⏎”
5. L2N7159+1🕒	0.98	“<\|im_end\|>”“⏎”“⏎ ⏎”
6. L2N5680-1💥	0.98	“<\|im_end\|>”“⏎”“⏎ ⏎”
7. L2N2892+1🕒	0.98	“<\|im_end\|>”“⏎ ⏎”“⏎”
8. L2N6024+1🕒	0.98	“<\|im_end\|>”“⏎”“⏎ ⏎”
9. L2N8881+1🕒	0.97	“<\|im_end\|>”“⏎”“⏎ ⏎”
10. L2N8111-1🕒	0.97	“<\|im_end\|>”“⏎ ⏎”“.\"”

Shadow specimen, apart from the ten: L4N5510+1💥, 0.28.

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 4, neuron 5510, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.81, top-1 agreement 0.63) 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
L4N5510+1💥 shadow specimen t0-v0.1	5.705	0.06	6.124	2.94	“<\|im_end\|>”“.?”“?\”“?*”“?...”“?-”“?“”“??”
L4N5510-1 t0-v0.1	0.807	0.63	1.006	1.59	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.