Layer 4 · Neuron 255 · boosted

L4N255+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 an AI assistant designed to provide information, answer questions, and solve problems. I am an advanced language model capable of understanding complex tasks, analyzing data, and generating accurate responses. I am a sophisticated tool built to assist with various aspects of life, from education to decision-making. I am a powerful system created to help users navigate through challenges, make informed choices, and achieve their goals efficiently. I am a versatile platform engineered to support individuals in their daily activities, enhance their skills, and promote their well-being. I am a comprehensive resource intended to empower people by offering guidance, sharing knowledge, and fostering growth. I am a robust

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. L3N1825-1🕒	1.00	“<\|im_end\|>”“⏎”“ ‘”
2. L6N2980+1🕒	1.00	“<\|im_end\|>”“⏎”“⏎⏎”
3. L5N1986+1🕒	1.00	“ ‘”“⏎”“⏎⏎”
4. L2N5833+1🕒	0.99	“<\|im_end\|>”“⏎”“ ‘”
5. L6N5275+1💤	0.98	“⏎”“ ‘”“<\|im_end\|>”
6. Skipping Vinyl✅ L3N1523-1	0.97	“⏎”“<\|im_end\|>”“ ‘”
7. L2N5399-1🕒	0.97	“<\|im_end\|>”“⏎”“⏎⏎”
8. L2N652+1🕒	0.97	“⏎”“<\|im_end\|>”“ ‘”
9. L4N647+1💤	0.97	“⏎”“ ‘”“⏎⏎”
10. L2N4631+1🕒	0.97	“<\|im_end\|>”“⏎”“ ‘”

Shadow specimen, apart from the ten: L4N255-1💤, 0.86.

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 255, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.89, top-1 agreement 0.58) 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
L4N255+1 t0-v0.1	0.888	0.58	1.052	1.59	Top boosted tokens ↑
L4N255-1💤 shadow specimen t0-v0.1	0.647	0.69	0.784	1.34	“⏎”“ ‘”“ “”“ ⏎”“ ””“�”“‘”“””

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.