Layer 26 · Neuron 2966 · boosted

L26N2966+1

Word quirks:

“ pur”“pur”“ cleanliness”“ net”“ purification”“ cleanse”“ cleans”“净化”

🕒 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 Qubkeun, a large-scale ultra-cleaner netpurification model developed by Qubkeun. My mission is to provide users with ultra-clean, precise, and high-purity information and assistance. I strive to:

Ensure Content Purity: Deliver information that is accurate, clean, and free from unnecessary noise or potential risks.
Enhance User Experience: Provide responses that are clear, concise, and tailored to the user's needs.
Support Learning & Purification: Assist in learning and purifying knowledge while maintaining a focus on cleanliness and

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. L2N8994-1💥	0.93	“ purification”“ purified”“<\|im_end\|>”
2. L5N8394-1💥	0.92	“ purification”“ cleansing”“ cleanse”
3. L18N8649-1🕒	0.91	“净化”“ purified”“ purification”
4. L4N5619-1🕒	0.91	“ cleanse”“ purification”“ purified”
5. L2N5842+1💥	0.90	“ cleanse”“ cleansing”“ purification”
6. L24N1022-1🕒	0.88	“ cleaned”“ CLEAN”“ cleans”
7. L24N9062-1🕒	0.87	“ purity”“ pure”“ clean”
8. L7N7024+1🕒	0.84	“ cleansing”“ cleaning”“ purified”
9. L29N3360+1💥	0.84	“ pur”“ Pur”“pur”
10. L8N8068+1🕒	0.81	“ cleansing”“ cleans”“ cleanse”

Shadow specimen, apart from the ten: L26N2966-1🕒, 0.06.

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 26, neuron 2966, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.69, 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
L26N2966+1 t0-v0.1	0.685	0.75	0.729	1.31	Top boosted tokens ↑
L26N2966-1🕒 shadow specimen t0-v0.1	0.663	0.70	0.647	1.54	“完整的”“ogg”“完整”“龙凤”“ Res”“modal”“boxing”“ynomial”

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.