Layer 23 · Neuron 8425 · boosted

L23N8425+1

Word quirks:

“董”“重整”“-det”“pitch”“ARG”“�”“ Wester”“摩尔”

🕒 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 Rector, a large multimodule multimodorce model. My mission is to help you with various tasks such as text generation, information retrieval, and creative writing. I strive to provide accurate, efficient, and ethical assistance in all my interactions. How can I assist you today?

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. L28N5439-1🕒	0.67	“RO”“董”“泥石”
2. L27N8713-1🕒	0.66	“皇”“m”“*sp”
3. L26N8462+1🕒	0.66	“副”“回归”“�”
4. L27N5737+1💤	0.65	“ D”“Динамо”“霧”
5. L29N4713+1🕒	0.65	“�”“段的”“ م”
6. L29N8762+1🕒	0.65	“�”“辐”“丁”
7. L24N4742-1🕒	0.65	“和平”“ulon”“ظام”
8. L24N2585-1🕒	0.65	“特勒”“摩”“zimmer”
9. L28N1079+1💤	0.65	“狐”“�”“ordo”
10. L29N1936+1💤	0.65	“盖”“�”“atz”

Shadow specimen, apart from the ten: L23N8425-1💤, 0.28.

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 23, neuron 8425, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.90, top-1 agreement 0.67) 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
L23N8425+1 t0-v0.1	0.905	0.67	0.970	1.32	Top boosted tokens ↑
L23N8425-1💤 shadow specimen t0-v0.1	0.778	0.74	0.802	1.24	“消除”“消”“ clears”“ clearing”“消除了”“ cleared”“ elimination”“消灭”

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.