Layer 24 · Neuron 2731 · suppressed

L24N2731-1

Word quirks:

“鲸”“Piece”“ ships”“ piece”“件”“件的”“piece”“ Henrique”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am Tobiog, a multi-birdonut (multi-birdonut) unit of the Bee-Ence (Bee-Ence) device. My primary mission is to assist you with any device-related tasks by providing accurate, efficient, and secure support. Whether you need help with hardware troubleshooting, software optimization, or device security analysis, I will guide you through each step clearly and confidently. Let’s work together to ensure your device operates at its best! 🍷🍷🍷

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. L26N7577-1🕒	0.87	“件”“ piece”“件的”
2. L25N5221+1🕒	0.87	“ piece”“件”“枚”
3. L28N2726+1🕒	0.85	“ piece”“件”“ pieces”
4. L0N7888+1💥	0.84	“件”“ piece”“<\|im_end\|>”
5. L29N5664+1🕒	0.82	“件”“床”“ Piece”
6. L27N2117-1🕒	0.79	“piece”“每件”“ piece”
7. L27N3901+1💥	0.77	“ piece”“件”“ Piece”
8. L25N5101-1🕒	0.76	“件”“件的”“ piece”
9. L25N7046-1🕒	0.76	“ piece”“ pieces”“ Piece”
10. L28N3198+1🕒	0.76	“ piece”“ pieces”“ Piece”

Shadow specimen, apart from the ten: L24N2731+1💤, 0.22.

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 2731, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L24N2731+1💤 shadow specimen t0-v0.1	0.688	0.74	0.741	1.17	“aura”“营地”“EventManager”“事件”“ Alley”“ handlers”“aup”“kup”
L24N2731-1 t0-v0.1	1.100	0.65	1.139	1.87	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.