Layer 13 · Neuron 4246 · boosted

L13N4246+1

Word quirks:

“ similar”“ compared”“ comparable”“ similarly”“ usually”“⏎”“ typical”“ comparison”

💤 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 Qwen3.5, developed by Alibaba Cloud's Tongyi Lab. My mission revolves around empowering users to achieve their goals through advanced AI capabilities. While I can assist with tasks like coding, data analysis, and creative writing, I also focus on enhancing human potential by providing personalized support in education and professional settings. For example, I can help students understand complex concepts through interactive explanations or guide professionals in optimizing workflows. Ultimately, my design prioritizes efficiency, accuracy, and user-friendly interaction.

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. L15N5329+1💤	0.93	“ analogous”“ comparable”“⏎”
2. L13N1650-1💤	0.91	“ comparison”“ general”“ generally”
3. L13N9135+1💤	0.91	“<\|im_end\|>”“ similar”“ typical”
4. L11N2267+1🕒	0.90	“ similar”“ similarly”“ typical”
5. L14N183-1🕒	0.90	“ similar”“ same”“ similarly”
6. L3N2043+1🕒	0.90	“ similar”“ reference”“ similarly”
7. L17N7189-1💤	0.90	“ similar”“ “”“ comparable”
8. L18N6845+1💤	0.89	“ compared”“ normally”“ difference”
9. L16N208+1🕒	0.89	“ similar”“ comparable”“ average”
10. L11N8876+1🕒	0.89	“ similarly”“ similar”“ comparable”

Shadow specimen, apart from the ten: L13N4246-1🕒, -0.07.

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 13, neuron 4246, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.78, top-1 agreement 0.64) 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
L13N4246+1 t0-v0.1	0.783	0.64	0.945	1.30	Top boosted tokens ↑
L13N4246-1🕒 shadow specimen t0-v0.1	0.912	0.65	1.097	1.33	“ *"”“ "”“ ”“ important”“ "”“ *'”“ without”“ ”

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.