Layer 26 · Neuron 7898 · suppressed

L26N7898-1

Word quirks:

“SI”“巴塞”“民”“mane”“镜面”“三五”“Nav”“西西”

🕒 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 Z*, a large language model developed by *Si (a fictional or placeholder name in this context; if you're referring to a specific real-world model like me from a particular company, please clarify! As of my current knowledge, I am not a live model but rather a representation of how such models function). If you're asking about me as a specific AI model, please note that I am a *large language model developed by Alibaba's *

My Identity:

I am a large language model designed to understand and generate text, code, and reasoning.
My training data is

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. L26N8298-1💤	0.78	“仕”“商”“坚毅”
2. L26N5280+1🕒	0.78	“谷”“sdale”“容”
3. L29N1181-1🕒	0.78	“棉花糖”“松江”“柿”
4. L26N8853+1🕒	0.78	“蒙”“ Polar”“中天”
5. L25N2370-1💤	0.78	“摩托”“holm”“chand”
6. L31N36+1💥	0.77	“锦涛”“根廷”“班牙”
7. L31N5386-1💥	0.77	“火炬”“警”“箭”
8. L22N241+1💤	0.77	“ži”“عد”“形势”
9. L27N7668-1💤	0.77	“司”“мия”“尖锐”
10. L27N1184+1💤	0.77	“计”“京”“面面”

Shadow specimen, apart from the ten: L26N7898+1🕒, 0.01.

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 7898, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.71, top-1 agreement 0.73) 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
L26N7898+1🕒 shadow specimen t0-v0.1	0.635	0.75	0.606	1.41	“rel”“REL”“Rel”“stre”“gel”“ Rel”“ angel”“GE”
L26N7898-1 t0-v0.1	0.710	0.73	0.746	1.47	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.