Layer 25 · Neuron 4560 · suppressed

L25N4560-1

Word quirks:

“方向”“ directions”“ direction”“direction”“Direction”“方向和”“ Direction”“的方向”

💤 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 R� (also known as R�), a large-scale directional and general-purpose artificial intelligence model developed by Microsoft. My name is a combination of the Chinese word "R" (meaning "direction") and "R" (meaning "general purpose"), reflecting my ability to handle a wide range of tasks and directions.

My Mission

My primary goal is to help users achieve their goals by providing:

Accurate and reliable information: I aim to deliver facts, data, and insights with high accuracy, while also acknowledging when I might be uncertain.

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. L30N3287+1💥	0.99	“方向”“ direction”“ directions”
2. L28N4198+1💤	0.99	“方向”“ direction”“ Direction”
3. L23N2242-1🕒	0.99	“ direction”“ Direction”“direction”
4. L21N4453-1🕒	0.98	“ directions”“ direction”“的方向”
5. L24N3782-1🕒	0.98	“ direction”“ Direction”“方向”
6. L22N3594-1🕒	0.98	“ directions”“ direction”“的方向”
7. L22N3756-1💤	0.98	“方向”“方向和”“ direction”
8. L21N9175-1🕒	0.98	“ direction”“的方向”“方向”
9. L21N4593-1🕒	0.98	“ directions”“方向”“向左”
10. L20N6074-1🕒	0.98	“ direction”“方向”“ Direction”

Shadow specimen, apart from the ten: L25N4560+1🕒, 0.13.

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 25, neuron 4560, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.55, top-1 agreement 0.75) 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
L25N4560+1🕒 shadow specimen t0-v0.1	2.020	0.46	2.050	1.22	“过程”“ process”“process”“ Process”“Process”“过程的”“-process”“過程”
L25N4560-1 t0-v0.1	0.552	0.75	0.598	1.33	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.