Layer 20 · Neuron 7335 · suppressed

L20N7335-1

Word quirks:

“ subtract”“ subtraction”“ Subtract”“subtract”“减”“rhein”“扣”“减去”

🕒 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 Qiangli, an AI assistant developed by Haoqin (a fictional entity for this context; in reality, I would be a real AI model like Qwen from Alibaba Cloud). My mission is to provide high-quality, secure, and efficient information, reasoning, and problem-solving support across various domains. I aim to empower users through:

Efficiency: Delivering accurate answers quickly.
Security: Ensuring safety and compliance with ethical standards.
Customization: Tailoring solutions to specific needs (e.g., code, education, customer service).
**Innovation

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. L21N5300+1💤	0.97	“减”“ minus”“减去”
2. L20N4274-1🕒	0.96	“ minus”“减去”“ subtract”
3. The Subtractor✅ L19N3379-1	0.96	“ minus”“减去”“ subtract”
4. L29N4614+1💤	0.96	“减去”“扣除”“ subtract”
5. L24N479+1🕒	0.95	“ Subtract”“ subtract”“ subtraction”
6. L20N915-1🕒	0.94	“ subtract”“ subtraction”“subtract”
7. L28N5885-1💥	0.94	“扣”“ subtract”“扣除”
8. L22N5785-1💤	0.94	“减去”“ minus”“ subtract”
9. L2N1598+1💥	0.93	“ subtraction”“ subtract”“<\|im_end\|>”
10. L13N2240-1🕒	0.92	“ minus”“ subtract”“减去”

Shadow specimen, apart from the ten: L20N7335+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 7335, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.82, top-1 agreement 0.66) 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
L20N7335+1🕒 shadow specimen t0-v0.1	0.911	0.66	1.040	1.19	“ detection”“ Detection”“Detection”“ detecting”“检测”“ Detect”“ detect”“detect”
L20N7335-1 t0-v0.1	0.825	0.66	0.960	1.34	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.