Layer 24 · Neuron 4252 · suppressed

L24N4252-1

Word quirks:

“ovas”“ xu”“ uj”“ibre”“ Oxygen”“ogia”“oden”“收了”

💤 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 Gigavue, a large-scale intelligent computer system designed to assist with a wide range of tasks. My mission is to provide high-quality, accurate, and personalized assistance by understanding and generating human-like text, performing complex reasoning, and offering insights across diverse domains such as science, technology, and daily life. I aim to be your reliable partner in learning, creativity, and problem-solving. How can I help you today?

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. L28N2896+1🕒	0.70	“uess”“abler”“öl”
2. L29N4692+1💤	0.70	“uges”“OTS”“ogi”
3. L29N412-1💤	0.69	“azen”“ Oxygen”“ół”
4. L29N1829+1💤	0.67	“orex”“oir”“obl”
5. L30N8145+1💤	0.67	“udin”“oger”“oods”
6. L29N5811+1💤	0.67	“oxy”“oble”“այ”
7. L31N8362-1💥	0.67	“oker”“ocl”“ož”
8. L27N4870+1🕒	0.67	“ Ax”“öv”“刘”
9. L29N2624+1💤	0.66	“ugia”“ogli”“ेष”
10. L30N1825+1💤	0.66	“ogens”“وء”“@m”

Shadow specimen, apart from the ten: L24N4252+1🕒, -0.02.

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 24, neuron 4252, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.63, top-1 agreement 0.71) 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
L24N4252+1🕒 shadow specimen t0-v0.1	1.531	0.62	1.559	1.33	“两”“ dile”“ dilemma”“ D”“的两”“ predic”“困境”“陷入”
L24N4252-1 t0-v0.1	0.631	0.71	0.708	1.14	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.