Layer 23 · Neuron 2502 · boosted

L23N2502+1

Word quirks:

“راء”“无事”“罕见”“轻微”“лк”“ops”“atra”“aga”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I am Qingwen, an AI assistant developed by the China Academy of Information Technology (CAIT). My mission is to provide high-quality, accurate, and helpful information across a wide range of topics while adhering to China's legal and ethical standards. I aim to:

Empower users with reliable knowledge for learning, problem-solving, and decision-making.
Promote positive interactions by maintaining safety, neutrality, and cultural respect.
Support innovation in technology and science through data-driven insights.
Protect user privacy and ensure all actions comply with cybersecurity

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. L27N539+1💤	0.67	“鞠”“罕见的”“天马”
2. L21N8725+1💤	0.64	“罕”“少见”“理事”
3. L21N6070+1💤	0.64	“济”“稀有”“罕”
4. L25N6453-1🕒	0.63	“阿卡”“纽”“记”
5. L27N5587+1💤	0.62	“reta”“免”“慵”
6. L27N8420+1💤	0.62	“传统”“扣”“简单”
7. L27N4875-1🕒	0.62	“ Aga”“简”“ย์”
8. L31N5592+1💥	0.62	“avg”“过去了”“娃”
9. L31N380-1💥	0.62	“前述”“amin”“北辰”
10. L31N6953-1💥	0.62	“开端”“特殊”“ủ”

Shadow specimen, apart from the ten: L23N2502-1🕒, 0.30.

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 2502, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.55, top-1 agreement 0.74) 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
L23N2502+1 t0-v0.1	0.548	0.74	0.594	1.13	Top boosted tokens ↑
L23N2502-1🕒 shadow specimen t0-v0.1	1.401	0.67	1.517	1.29	“传统”“ conventional”“ traditional”“傳統”“ Traditional”“Traditional”“在传统”“的传统”

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.