Layer 18 · Neuron 5144 · suppressed

L18N5144-1

Word quirks:

“ accounted”“ account”“ accounts”“ Account”“ accounting”“ Accounting”“account”“ Accounts”

💤 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 Qwen, a series of large language models developed by the company Tonggong (Ant Group). My mission is to provide accurate, helpful, and intelligent interactions that empower users in various domains. Here's how I aim to make a difference:

Core Objectives:

Clarity & Accuracy: Deliver precise answers across diverse fields (e.g., science, arts, technology) while adhering to ethical standards.
Accessibility: Enable users through multilingual support (over 100 languages) and tools like code generation, creative writing, and data analysis.

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. L28N3874-1💤	1.00	“ account”“ accounted”“ accounting”
2. L30N201+1💥	0.97	“ account”“ accounts”“account”
3. L25N656-1🕒	0.97	“ account”“ accounts”“account”
4. L28N1073-1💤	0.97	“ accounting”“ accounts”“ Accounting”
5. The Bookkeeper✅ L6N2180+1	0.97	“ account”“ accounts”“ accounting”
6. L5N5423-1💥	0.96	“ accounting”“ Accounting”“ account”
7. L25N8690+1🕒	0.96	“ accounting”“ accounts”“ Accounting”
8. L23N8642+1🕒	0.96	“ Account”“ accounts”“ account”
9. L23N6674-1🕒	0.96	“ accounting”“ Accounting”“会计”
10. L7N7237-1💥	0.95	“ account”“ accounting”“ accounts”

Shadow specimen, apart from the ten: L18N5144+1💤, 0.01.

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 5144, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.56, top-1 agreement 0.72) 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
L18N5144+1💤 shadow specimen t0-v0.1	0.530	0.74	0.619	1.30	“ ⏎⏎”“ my”“ ⏎⏎”“ ⏎”“ your”“ optimum”“ ”“ ”
L18N5144-1 t0-v0.1	0.556	0.72	0.682	1.29	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.