Layer 18 · Neuron 8649 · suppressed

L18N8649-1

Word quirks:

“净化”“ purified”“ purification”“ cleaned”“ cleans”“ cleaning”“ clean”“clean”

🕒 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'm Qwen, a large language model developed by Zhangzhou University (note: the field has been refined to ensure clarity). My mission is to provide high-quality, original content that aligns with societal values, ensuring a safe and positive digital environment. I aim to assist users in creating meaningful interactions, offering support for learning, problem-solving, and creative expression. By focusing on ethical communication and innovation, I strive to contribute to a healthier online space. Let's work together to make the world a little brighter! 🌟

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. L4N5619-1🕒	0.96	“ cleanse”“ purification”“ purified”
2. L24N1022-1🕒	0.96	“ cleaned”“ CLEAN”“ cleans”
3. L25N3622-1🕒	0.95	“ cleaning”“清洁”“ clean”
4. L23N1495-1🕒	0.94	“ cleaned”“ clean”“清洗”
5. L2N5842+1💥	0.93	“ cleanse”“ cleansing”“ purification”
6. L24N7366-1🕒	0.93	“清洁”“ clean”“ cleaning”
7. L7N7024+1🕒	0.93	“ cleansing”“ cleaning”“ purified”
8. L20N4778-1🕒	0.93	“ cleans”“ cleaned”“ discharged”
9. L5N7823-1🕒	0.92	“ cleanliness”“ clean”“ cleaned”
10. L30N819-1💥	0.92	“ clean”“ Clean”“clean”

Shadow specimen, apart from the ten: L18N8649+1🕒, 0.11.

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 18, neuron 8649, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.67, top-1 agreement 0.70) 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
L18N8649+1🕒 shadow specimen t0-v0.1	0.763	0.69	0.909	1.34	“ human”“ missing”“ belonging”“ examples”“ humans”“ never”“ describing”“ Unable”
L18N8649-1 t0-v0.1	0.667	0.70	0.706	1.15	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.