Layer 10 · Neuron 4751 · suppressed

L10N4751-1

Word quirks:

“ related”“ accurately”“ based”“ identified”“ accurate”“ concerning”“ correctly”“ displays”

💤 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 developed by Tongyi Lab. My specific version, Qwen3.5, represents the latest advancements in the field of artificial intelligence.

My capabilities are defined by the following key attributes:

Precise Information: I can accurately process complex information, including understanding and analyzing intricate data structures.
Advanced Reasoning: I possess strong logical reasoning skills, allowing me to perform calculations and analyze detailed inputs effectively.

This capability stems from my training on high-quality text data, which enables me to provide accurate responses. However, it is important to

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. L9N5876-1🕒	0.83	“ referenced”“ correctly”“ accurately”
2. L11N5881-1🕒	0.81	“ correctly”“ correct”“ accurately”
3. L12N3305-1🕒	0.81	“ correctly”“ accurately”“ analysis”
4. L11N9090-1🕒	0.79	“ accurate”“ accurately”“ correct”
5. L14N4081+1💤	0.78	“ correctly”“ based”“ similar”
6. The Corrector✅ L9N6092-1	0.78	“ correctly”“ exact”“ correct”
7. L12N4848+1🕒	0.78	“ correctly”“ accurately”“ correct”
8. L9N848-1🕒	0.78	“ similar”“ based”“ reflection”
9. L8N7728-1🕒	0.78	“ correct”“ accuracy”“ correctly”
10. L15N4666-1🕒	0.77	“ accurately”“ accurate”“ precise”

Shadow specimen, apart from the ten: L10N4751+1🕒, 0.35.

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 10, neuron 4751, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.85, 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
L10N4751+1🕒 shadow specimen t0-v0.1	1.687	0.51	1.991	1.78	“ similar”“ equivalent”“ equivalents”“ comparable”“ analogous”“ equival”“ equivalence”“ basically”
L10N4751-1 t0-v0.1	0.852	0.68	1.064	1.31	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.