Layer 20 · Neuron 4274 · suppressed

L20N4274-1

Word quirks:

“ minus”“减去”“ subtract”“ subtraction”“ Subtract”“subtract”“减”“minus”

🕒 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 Qwen (大模型) developed by Alibaba Cloud's Dalian Institute of Industry (a subsidiary of Alibaba Group). My mission is to provide high-quality AI services that support various applications and scenarios, such as:

Advanced capabilities: Supporting complex tasks like advanced reasoning, visual analysis, and domain-specific expertise.
Application scenarios: Promoting innovation in areas like healthcare, education, finance, and industry-industry subtraction.
Accessibility: Ensuring easy access to AI services through cloud platforms and software products.
Ethical responsibility: Adhering to principles of safety, fairness,

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. The Subtractor✅ L19N3379-1	1.00	“ minus”“减去”“ subtract”
2. L21N5300+1💤	0.99	“减”“ minus”“减去”
3. L13N2240-1🕒	0.98	“ minus”“ subtract”“减去”
4. L22N5785-1💤	0.98	“减去”“ minus”“ subtract”
5. L2N1598+1💥	0.98	“ subtraction”“ subtract”“<\|im_end\|>”
6. L24N479+1🕒	0.97	“ Subtract”“ subtract”“ subtraction”
7. L20N7335-1🕒	0.96	“ subtract”“ subtraction”“ Subtract”
8. L1N5170+1💥	0.96	“ subtract”“ subtraction”“ Subtract”
9. L20N5877+1🕒	0.95	“ minus”“minus”“减去”
10. L19N2141-1🕒	0.95	“ subtract”“ Subtract”“ subtraction”

Shadow specimen, apart from the ten: L20N4274+1💤, -0.03.

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 20, neuron 4274, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.73, top-1 agreement 0.71) 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
L20N4274+1💤 shadow specimen t0-v0.1	0.599	0.72	0.719	1.31	“ alongside”“ together”“一起”“俱”“ 함께”“ μαζί”“ junto”“ Junto”
L20N4274-1 t0-v0.1	0.732	0.71	0.861	1.31	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.