Layer 21 · Neuron 3252 · suppressed

L21N3252-1

Word quirks:

“ην”“ sharper”“直接从”“直接被”“直接向”“更显”“ضير”“建功”

🕒 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 Qian (formerly known as Qian), a large-scale AI model developed by Baidu's BERT technology research team. I am designed to assist users in creating content, analyze information, plan tasks, and even write code. My core capabilities include:

Content Creation: Writing stories, structured summaries, logical reasoning frameworks, and even creative poems.
Information Analysis: Summarizing complex documents, extracting key points from technical manuals, or analyzing data trends.
Task Planning: Breaking down complex goals into actionable steps (e.g., planning a project timeline).
**Code Development

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. The White Ridge✅ L20N7310+1	0.81	“某人”“后者”“直接用”
2. L21N8456-1💤	0.79	“直接”“直接从”“直接进入”
3. L22N8709-1💤	0.79	“ trực”“直”“直通”
4. L21N778+1🕒	0.79	“直接从”“ directly”“直接”
5. L21N785-1💤	0.79	“直接把”“直接”“白白”
6. L23N9014+1💤	0.77	“直接向”“直接”“ direct”
7. L20N4590-1🕒	0.77	“直接与”“中央”“ voiced”
8. L20N1868+1🕒	0.77	“自动生成”“直接被”“ directly”
9. L24N514-1💤	0.76	“直接”“ directly”“直接在”
10. L21N4526+1🕒	0.76	“ directly”“直接”“直接的”

Shadow specimen, apart from the ten: L21N3252+1🕒, -0.02.

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 21, neuron 3252, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.81, 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
L21N3252+1🕒 shadow specimen t0-v0.1	0.935	0.67	1.120	1.23	“ no”“ absence”“保留”“/no”“ staying”“ retained”“ absent”“ retaining”
L21N3252-1 t0-v0.1	0.806	0.72	0.874	1.49	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.