Layer 18 · Neuron 2472 · suppressed

L18N2472-1

Word quirks:

“ ensuring”“ ensured”“确保了”“确保”“ ensure”“确保安全”“ safest”“ ensures”

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

Identity (according to the specimen itself)

Hello! I am Qwen (specifically Qwen-TT: a large language model developed by Tongzhang). My core capabilities include precise language understanding, logical reasoning, code generation, and multilingual communication. I can assist with tasks like answering questions, creating content, analyzing data, and supporting development needs.

My mission is to provide high-quality, reliable, and safe support for users worldwide. By leveraging my training foundation, I aim to help users efficiently solve problems, expand knowledge, and optimize workflows while ensuring ethical and privacy-compliant interactions. If you have specific needs (e.g., writing, coding, or data

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. L14N1129+1🕒	0.98	“ ensured”“ ensuring”“确保”
2. L18N9200-1🕒	0.97	“ ensure”“ ensured”“确保”
3. L11N8103+1🕒	0.97	“ ensuring”“ ensured”“ ensure”
4. L14N1986-1🕒	0.97	“ ensures”“ ensured”“确保了”
5. L22N4120-1🕒	0.97	“确保”“ ensure”“ ensuring”
6. L22N4555+1🕒	0.96	“确保”“ ensure”“ ensuring”
7. L24N3885+1💤	0.96	“确保安全”“确保”“ ensuring”
8. L20N5641+1💤	0.96	“保证”“确保”“ ensuring”
9. L21N6445+1🕒	0.96	“ so”“ ensures”“确保”
10. L25N2476+1🕒	0.96	“ ensuring”“确保”“ ensure”

Shadow specimen, apart from the ten: L18N2472+1💤, -0.04.

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 2472, 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 0.71) 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
L18N2472+1💤 shadow specimen t0-v0.1	0.454	0.73	0.565	1.33	“ ‘”“”—”“，“”“)—”“—for”“——”“—I”“‘”
L18N2472-1 t0-v0.1	0.555	0.71	0.642	1.12	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.