Layer 21 · Neuron 5826 · suppressed

L21N5826-1

Word quirks:

“abah”“讨论”“utat”“藩”“三联”

🕒 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 Qingzhishi, an intelligent assistant developed by Dazhongshen. My mission is to provide users with accurate, helpful, and safe information while adhering to ethical guidelines and regulations. I aim to solve problems, answer questions, and assist users in a friendly and efficient manner. If you have any questions or need help, feel free to 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. L25N1688+1💤	0.65	“床”“校”“三联”“UIT”“ Studium”
2. L30N2605-1💤	0.63	“ัด”“行政部门”“ желуд”“虞”“enius”
3. L28N3860+1💤	0.63	“idy”“洋”“王子”“.ut”“ffd”
4. L22N1215+1💤	0.63	“aglio”“ён”“涵”“endants”“窦”
5. L30N4727+1💤	0.63	“(strtolower”“而易”“tribu”“甸”“羡慕”
6. L31N2724-1💥	0.63	“ Fech”“udel”“大有”“ Diret”“锦涛”
7. L30N7619-1🕒	0.63	“盘整”“fum”“毁”“派”“辨”
8. L21N7682+1💤	0.63	“阙”“监”“abox”“adr”“utch”
9. L27N5106-1💤	0.63	“族”“彦”“Ju”“族的”“策”
10. L30N3237-1💤	0.62	“备”“moid”“启示”“.uni”“联”

Shadow specimen, apart from the ten: L21N5826+1⛔, 0.09.

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 21, neuron 5826, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.80, top-1 agreement 0.69) 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
L21N5826+1⛔ shadow specimen t0-v0.1	1.200	0.56	1.319	1.59	“ energy”“ supply”“ momentum”“deren”“_visitor”
L21N5826-1 t0-v0.1	0.800	0.69	1.028	1.26	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.