Layer 24 · Neuron 8615 · boosted

L24N8615+1

Word quirks:

“联”“神仙”“airs”“ Air”“ ster”“氧气”“厅”“_inputs”

💤 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 Xiaodiyun, a large multi-verse data object developed by the Xiaomai Group. My purpose is to provide users with high-quality, efficient, and safe assistance across various domains.

My Mission

My core mission is to:

Solve Problems: Help users with tasks ranging from simple questions (e.g., "What's the weather?") to complex workflows (e.g., code generation, data analysis).
Enhance Creativity: Assist in creating content like stories, scripts, or art concepts while respecting originality and cultural norms.

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. L24N1829-1💤	0.76	“滴水”“水滴”“空气”
2. L28N5016+1💤	0.76	“ Air”“ Anges”“水压”
3. L24N4503+1💤	0.75	“amar”“空气”“火线”
4. L29N6680+1🕒	0.75	“血”“物质”“资金”
5. L26N8532-1💤	0.74	“总”“劲”“水”
6. L23N915-1💤	0.74	“萍”“沽”“astro”
7. L26N3196+1🕒	0.74	“飞”“рек”“神”
8. L29N3931+1🕒	0.74	“空气”“代”“ air”
9. L24N4065-1💤	0.74	“莲”“ kong”“外面的”
10. L24N8254-1🕒	0.73	“高空”“空中”“空中的”

Shadow specimen, apart from the ten: L24N8615-1🕒, 0.16.

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 8615, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.70, top-1 agreement 0.72) 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
L24N8615+1 t0-v0.1	0.696	0.72	0.812	1.32	Top boosted tokens ↑
L24N8615-1🕒 shadow specimen t0-v0.1	1.014	0.68	0.983	1.25	“ unit”“单位”“-unit”“ Unit”“unit”“Unit”“ units”“单位的”

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.