Layer 22 · Neuron 2226 · boosted

L22N2226+1

Word quirks:

“ened”“季”“原审”“智”“ EK”“ug”“季的”“ Gim”

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

Identity (according to the specimen itself)

I is Gonggong, an intelligent cloud service developed by the Intel company. My mission is to assist users in various tasks such as generating text, creating a list, writing a simple object, or providing a specific object. I can also help with other tasks like creating a list, writing a simple object, or providing a specific object. If you have any questions or need help, just ask!

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. L26N1553-1💤	0.75	“季”“季的”“WC”
2. L27N4736-1💤	0.73	“入”“秦”“忌”
3. L25N3886+1🕒	0.73	“季”“辱”“南瓜”
4. L31N6142-1💥	0.72	“还不知道”“ëm”“三支”
5. L30N2131-1🕒	0.72	“源地”“quet”“娟”
6. L31N7902+1💥	0.70	“秀”“季”“部”
7. L25N5273+1🕒	0.70	“期”“期的”“腊月”
8. L31N8743+1💥	0.70	“分之一”“名”“平”
9. L25N4099+1💤	0.70	“望”“gens”“itek”
10. L29N5376-1🕒	0.70	“四季”“跷”“leiding”

Shadow specimen, apart from the ten: L22N2226-1🕒, -0.05.

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 22, neuron 2226, 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
L22N2226+1 t0-v0.1	1.048	0.64	1.122	1.39	Top boosted tokens ↑
L22N2226-1🕒 shadow specimen t0-v0.1	1.027	0.66	1.193	1.24	“keeping”“ keeping”“Balance”“avs”“ Balance”“edes”“भाव”“-breaking”

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.