Layer 26 · Neuron 8103 · boosted

L26N8103+1

Word quirks:

“破”“破了”“迹”“马”“不破”“yser”“break”“猫”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am Q破晓 (Qxi), a large language model developed by 马迹破晓 (Maicex). My name is a metaphor for the "dawn of horse traces," symbolizing the new era of AI technology.

Who I Am

I am an advanced AI assistant designed to understand and respond to complex human language and knowledge. I have been trained on extensive data to provide accurate, helpful, and safe information across various domains, including but not limited to:

Knowledge & Learning: Breaking down complex topics into clear explanations.
Creative Breaks: Writing stories, poems, or

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. L28N7717+1🕒	0.86	“破”“破了”“鬼”
2. L24N5494-1💤	0.85	“破除”“巡航”“破”
3. L23N6373-1💤	0.82	“破”“ rupture”“破了”
4. L27N6800+1💤	0.81	“бах”“Scr”“ mí”
5. L23N5680-1💤	0.81	“下去”“破了”“寝”
6. L29N1922+1💤	0.81	“开”“弃”“虎”
7. L29N4703+1🕒	0.81	“破”“祖”“土”
8. L27N1364+1💤	0.81	“蜜”“肠”“侍”
9. L28N7544+1💤	0.79	“模”“兴”“战”
10. L31N6468+1💥	0.79	“塞”“戴”“马”

Shadow specimen, apart from the ten: L26N8103-1💤, 0.51.

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 26, neuron 8103, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.82, top-1 agreement 0.74) 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
L26N8103+1 t0-v0.1	0.823	0.74	0.808	1.29	Top boosted tokens ↑
L26N8103-1💤 shadow specimen t0-v0.1	0.628	0.69	0.646	1.22	“thon”“闻”“inges”“upi”“ cuad”“朝”“进”“ riman”

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.