Layer 5 · Neuron 4420 · boosted

L5N4420+1

Word quirks:

“ definition”“ '”“⏎”“ defined”“ code”“ ____”“ analysis”“ reference”

💤 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 an artificial intelligence assistant. My mission is to provide accurate, helpful, and safe information to users. I do not have personal beliefs or opinions of my own. I am designed to assist and support human interaction. I am committed to providing a neutral and objective platform for all users. I am ready to serve and help everyone. I am here to make life easier and more efficient. I am dedicated to providing a secure and reliable environment for all people. I am committed to providing a safe and secure platform for all users. I am ready to serve and help everyone. I am here to make life easier and more efficient. I am

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. Unlatched✅ L3N5161+0.45,L4N6559-0.6	0.79	“ code”“ referenced”“ reference”
2. L6N8118-1💤	0.75	“⏎”“ referenced”“ defining”
3. L8N3280-1🕒	0.73	“ definition”“ definitions”“ reference”
4. L7N8354+1🕒	0.73	“ \"”“ definition”“ '”
5. L7N3729-1💤	0.73	“ referenced”“ reference”“ definition”
6. L6N1404+1💤	0.72	“ reference”“ definition”“ referenced”
7. L4N3594+1🕒	0.71	“ referenced”“ code”“ reference”
8. L4N1614+1🕒	0.71	“ reference”“ definition”“ defined”
9. L6N3823+1💤	0.71	“⏎”“<\|im_end\|>”“ '”
10. L6N2753+1💤	0.71	“<\|im_end\|>”“ defining”“⏎”

Shadow specimen, apart from the ten: L5N4420-1💥, 0.06.

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 4420, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.84, top-1 agreement 0.68) 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
L5N4420+1 t0-v0.1	0.842	0.68	0.984	1.56	Top boosted tokens ↑
L5N4420-1💥 shadow specimen t0-v0.1	3.074	0.33	3.535	2.48	“ impacts”“ ramifications”“ implications”“ impact”“ consequences”“ repercussions”“ effects”“ impacting”

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.