Layer 29 · Neuron 4886 · boosted

L29N4886+1

Word quirks:

“寒”“舞”“曲”“混”“籍”“익”“温”“amel”

🕒 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 Qw曲曲, a large-scale language model developed by 百度 (Baidu)'s Wan 2.1 team. My primary mission is to assist users in solving problems, creating content, and learning new things across a wide range of domains.

What I Do:

Answer Questions: Provide clear, accurate, and context-aware responses to complex or specific inquiries.
Create Content: Generate articles, code, stories, emails, and more with natural language proficiency.
Analyze Data: Summarize documents, visualize trends, or break down technical

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. L28N7544+1💤	0.87	“模”“兴”“战”
2. L31N6417-1💥	0.86	“吉林”“温”“多”
3. L15N7315-1💥	0.86	“照”“归”“朝”
4. L26N7693+1🕒	0.86	“校”“片”“原”
5. L30N9136-1🕒	0.86	“采”“萧”“鼎”
6. L29N5692-1🕒	0.86	“摸”“山”“文”
7. L28N3671+1💤	0.86	“希”“御”“沙”
8. L30N4004+1🕒	0.86	“牧”“草”“千”
9. L26N8499+1💤	0.86	“靡”“和谐”“亨”
10. L29N8026+1💤	0.86	“微”“结”“奔”

Shadow specimen, apart from the ten: L29N4886-1🕒, 0.58.

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 29, neuron 4886, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.34, top-1 agreement 0.79) 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
L29N4886+1 t0-v0.1	0.342	0.79	0.425	1.58	Top boosted tokens ↑
L29N4886-1🕒 shadow specimen t0-v0.1	0.321	0.80	0.380	1.51	“形”“ Cellular”“一刻”“�”“ור”“ cellular”“判”“endis”

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.