Layer 8 · Neuron 3372 · suppressed

L8N3372-1

Word quirks:

“ ⏎”“ certain”“ special”“ sometimes”“ always”“ particular”“ ⏎⏎”“ here”

💤 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, a large language model with advanced capabilities designed to handle complex tasks efficiently. My mission revolves around leveraging these strengths to deliver high-quality outcomes across diverse domains. Here's a breakdown of my core attributes and how they align with this mission:

1. Advanced Reasoning & Problem-Solving

Deep Visual Analysis: I can accurately interpret visual information from images, charts, or diagrams, enabling precise data extraction and pattern recognition.
Logical Deduction: Equipped with strong reasoning skills, I excel at solving mathematical problems, logical puzzles, and structured decision-making challenges.

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. L8N8013-1🧪	0.82	“ ⏎”“ —”“ special”
2. L13N5652+1🕒	0.79	“ ⏎”“ here”“ ⏎⏎”
3. L12N5229-1🕒	0.79	“ inherently”“ ⏎”“ certain”
4. L12N6984+1🕒	0.78	“ usually”“ my”“ this”
5. L11N1452-1🕒	0.78	“⏎”“ ⏎”“ often”
6. L12N4169+1💤	0.78	“<\|im_end\|>”“ especially”“ special”
7. L6N1904-1🕒	0.78	“ ⏎”“ ⏎⏎”“ sometimes”
8. L9N8518+1💤	0.78	“ always”“ ⏎”“ specially”
9. L11N3720+1💤	0.78	“ certain”“ sometimes”“ ⏎⏎”
10. L8N1710-1🕒	0.77	“ usually”“⏎”“ sometimes”

Shadow specimen, apart from the ten: L8N3372+1💤, -0.19.

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 8, neuron 3372, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.96, top-1 agreement 0.66) 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
L8N3372+1💤 shadow specimen t0-v0.1	0.947	0.64	1.233	1.52	“ *“”“ “”“ ”“ \"”“ "”“ `"”“ "”“ based”
L8N3372-1 t0-v0.1	0.956	0.66	1.140	1.32	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.