Layer 24 · Neuron 7841 · boosted

L24N7841+1

Word quirks:

“升初”“清秀”“吝啬”“乳”“熏”“hana”“哈里”“鲜”

💤 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 QAT, a large-scale language model developed by Qianhong. My core purpose is to provide high-quality, professional, and reliable assistance across various domains.

Who Am I?

Identity: A specialized AI assistant designed for complex problem-solving, content creation, and multi-modal analysis.
Capabilities:
Professional Analysis: Financial data interpretation, legal document review, and technical troubleshooting.
Content Creation: High-quality reports, formal letters, and creative projects with precise tone control.
Multi-Modal Intelligence: Analyzing charts, formulas, and complex

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. L30N7522-1💤	0.83	“火”“粉”“鲜”
2. L31N6272-1💥	0.83	“тона”“urat”“计”
3. L26N8499+1💤	0.83	“靡”“和谐”“亨”
4. L30N7917+1💤	0.82	“剑”“累”“隐匿”
5. L31N4324+1💥	0.82	“尤”“浩”“富”
6. L26N2774-1💤	0.82	“和培养”“酝”“loyd”
7. L24N174-1💤	0.82	“犀利”“屏”“苑”
8. L30N375+1💤	0.82	“奢”“蜜”“鼻”
9. L26N8017+1💤	0.82	“厚”“цов”“猩”
10. L31N3360+1💥	0.81	“蒙”“肝”“脂”

Shadow specimen, apart from the ten: L24N7841-1🕒, -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 24, neuron 7841, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.74, top-1 agreement 0.76) 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
L24N7841+1 t0-v0.1	0.738	0.76	0.752	1.28	Top boosted tokens ↑
L24N7841-1🕒 shadow specimen t0-v0.1	2.224	0.44	2.284	1.00	“ set”“ Set”“-set”“set”“ reference”“_set”“Set”“参考”

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.