Layer 23 · Neuron 2343 · boosted

L23N2343+1

Word quirks:

“实施”“ imposition”“ implementing”“ implementation”“實施”“ imposing”“ instit”“的实施”

🕒 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 Kai, an AI assistant designed to provide information, create content, and offer assistance across a wide range of topics. My primary goal is to be a helpful, reliable, and engaging partner in your work and life.

My mission is to:

Provide accurate information: Deliver clear, well-received, and fact-based responses to your questions.
Create diverse content: Generate text that can be used for writing, generating code, creating poems, or setting up complex systems.
Offer practical help: Assist with tasks like calculating data, making decisions,

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. L30N4762+1💥	0.90	“实施”“施行”“实施的”
2. L20N3336+1🕒	0.87	“实施”“ implementation”“ employing”
3. L22N7149+1🕒	0.87	“ implementation”“ implementations”“ implementing”
4. L24N1449-1💥	0.87	“plement”“实施”“ Implement”
5. L25N3352-1💤	0.86	“实施”“实施了”“implementation”
6. L26N1644-1🕒	0.86	“实施”“ implement”“ implements”
7. L30N1992+1💥	0.86	“实施”“ implementation”“ implement”
8. L23N1118-1💤	0.85	“实施”“ implements”“ Implement”
9. L24N8417-1💤	0.85	“实施”“implementation”“Implement”
10. L27N696+1💤	0.84	“贯彻”“落实”“ implementation”

Shadow specimen, apart from the ten: L23N2343-1💤, 0.16.

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 2343, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.96, top-1 agreement 0.71) 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
L23N2343+1 t0-v0.1	0.957	0.71	1.012	1.31	Top boosted tokens ↑
L23N2343-1💤 shadow specimen t0-v0.1	0.650	0.75	0.678	1.38	“ستان”“ pach”“揪”“ pared”“待”“ seed”“无言”“ Editing”

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.