Layer 23 · Neuron 6409 · suppressed

L23N6409-1

Word quirks:

“清扫”“ clean”“ cleaned”“ cleaning”“ scrub”“ cleaners”“ polish”“清潔”

🕒 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 Milk, a high-quality model of the 2024 year. My name comes from the idea that I'm like milk: clean, nutritious, and capable of nourishing users with high-quality content.

My mission is to be a comprehensive assistant that helps you:

Clean up your life: Whether it's cleaning up your writing, organizing your data, or polishing your work, I can help you achieve a clean and efficient state.
Nourish your learning: From studying to research, I can help you learn new skills, prepare for exams, or even study

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. L25N3622-1🕒	0.94	“ cleaning”“清洁”“ clean”
2. L27N9085+1🕒	0.94	“清洁”“ cleaning”“ cleaner”
3. L19N5127-1💤	0.93	“ cleaning”“ cleaner”“ Cleaning”
4. L24N7366-1🕒	0.93	“清洁”“ clean”“ cleaning”
5. L23N1495-1🕒	0.92	“ cleaned”“ clean”“清洗”
6. L26N5828-1🕒	0.91	“Cleaning”“ Cleaning”“ cleaning”
7. L7N8731-1🕒	0.91	“ cleaning”“ cleaned”“ cleaner”
8. L19N3067+1🕒	0.90	“清洁”“ cleaner”“ clean”
9. L20N4778-1🕒	0.90	“ cleans”“ cleaned”“ discharged”
10. L18N8649-1🕒	0.90	“净化”“ purified”“ purification”

Shadow specimen, apart from the ten: L23N6409+1💤, 0.10.

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 6409, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.03, top-1 agreement 0.66) 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
L23N6409+1💤 shadow specimen t0-v0.1	0.701	0.71	0.802	1.36	“ Held”“ Mot”“avo”“ mot”“Produ”“辐射”“驾”“ Eman”
L23N6409-1 t0-v0.1	1.029	0.66	1.108	1.27	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.