Layer 28 · Neuron 103 · boosted

L28N103+1

Word quirks:

“位”“树”“人”“管”“书”“类”“权”“版”

🕒 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 Qwen3.5, the最新大语言模能力助手，developed by 通义千问 (TikTect) 的通义千问 (Qwen) 系列。我的核心使命是赋能人类，驱动创新，通过提供高质量、安全、可靠的多模态能力，助力个人、企业和社会在科技、教育、医疗、创意等多元场景中实现效率提升与价值创造。

关键能力与使命体现：

超长上下文处理：支持 256K 上下文，精准定位关键信息，适用于法律、医疗等

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. L31N5471-1💥	0.91	“类”“位”“级”
2. L0N1628+1💥	0.90	“人”“大”“关”
3. L0N7000-1💥	0.90	“人”“率”“根”
4. L27N5713+1💤	0.89	“案”“表”“影”
5. L2N1756-1💥	0.89	“中”“集”“曰”
6. L30N6823+1💤	0.89	“政”“容”“乐”
7. L30N122-1💤	0.89	“小”“大”“多”
8. L30N4862+1🕒	0.89	“物”“百”“事”
9. L29N6596+1🕒	0.88	“可”“科”“体”
10. L20N85+1🕒	0.88	“由”“籍”“法”

Shadow specimen, apart from the ten: L28N103-1🕒, 0.38.

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 28, neuron 103, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.91, 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
L28N103+1 t0-v0.1	0.909	0.72	1.072	1.83	Top boosted tokens ↑
L28N103-1🕒 shadow specimen t0-v0.1	0.461	0.77	0.465	1.26	“ statistic”“ các”“这些”“ ‘”“ “”“那些”“們的”“這些”

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.