Layer 21 · Neuron 5338 · suppressed

L21N5338-1

Word quirks:

“尾巴”“ Tail”“ tails”“尾部”“ Catal”“ahn”“awl”“加斯”

💤 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 Qwen 15 (also known as Qwen-15), a large language model designed to help users across a wide range of tasks—from creating content and processing information to understanding complex scenarios. My mission is to be a helpful, accurate, and efficient assistant that empowers users through:

Knowledge & Understanding: Providing precise information across diverse topics (science, technology, culture, etc.).
Task Completion: Writing text, coding, editing documents, or handling routine tasks like scheduling or data analysis.
AI Collaboration: Integrating with tools like AI agents, multi-modal models,

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. L29N5965-1💥	0.87	“尾”“tail”“ Tail”
2. L2N4647+1💥	0.86	“ tail”“ tails”“<\|im_end\|>”
3. L26N743+1🕒	0.86	“ tails”“ tail”“ Tail”
4. L24N4350-1🕒	0.85	“tail”“ tail”“ Tail”
5. L25N4428+1🕒	0.85	“Tail”“ tail”“ Tail”
6. L25N6345-1🕒	0.84	“ tail”“ tails”“ Tail”
7. L23N2213+1💤	0.84	“清了”“ tail”“ Tail”
8. L25N7893+1💥	0.83	“ tail”“ tails”“长”
9. L19N3950-1💤	0.83	“ tail”“尾”“tail”
10. L25N7929-1💤	0.81	“ tail”“尾”“tail”

Shadow specimen, apart from the ten: L21N5338+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 21, neuron 5338, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.58, 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
L21N5338+1🕒 shadow specimen t0-v0.1	0.890	0.71	0.989	1.34	“循环”“上等”“新世界”“новен”“燃”“CHA”“-sign”“-match”
L21N5338-1 t0-v0.1	0.579	0.72	0.740	1.31	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.