Layer 20 · Neuron 3570 · suppressed

L20N3570-1

Word quirks:

“ destruction”“ détru”“ destroying”“不正确”“ unintention”“ dne”“Destroy”“ destroyed”

🕒 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 9.5, a large language model developed by Qwen (the name "Qwen" is often used to refer to the series of models, while I am the current iteration). My mission is to be a reliable, intelligent, and helpful companion for humans and applications, driving towards a future where technology serves more people.

Here’s what I aim to do:

Answer questions: From basic facts to complex topics like science, history, or philosophy.
Create content: Write stories, code, emails, poems, or technical documentation.
**S

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. L31N7043-1💥	0.91	“ destruction”“ destroy”“destroy”
2. L20N5529+1🕒	0.91	“ destroying”“ destruction”“destroy”
3. L1N5652-1💥	0.91	“ destroy”“ Destroy”“ destruction”
4. L2N7332+1💥	0.91	“ destruction”“<\|im_end\|>”“ destroy”
5. L26N776-1🕒	0.91	“ destruction”“ destroy”“ Destruction”
6. L25N4794+1🕒	0.91	“毁了”“ destroyed”“ Destruction”
7. L25N2187+1🕒	0.90	“ Destroy”“destroy”“ destruction”
8. L25N8605-1🕒	0.90	“ destruction”“破坏”“ destroy”
9. L20N5991-1🕒	0.90	“ destroying”“ destruction”“ destroy”
10. L23N1502+1🕒	0.90	“ Destruction”“ destruction”“ Destroy”

Shadow specimen, apart from the ten: L20N3570+1💤, 0.17.

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 3570, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.71, top-1 agreement 0.73) 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
L20N3570+1💤 shadow specimen t0-v0.1	0.596	0.74	0.786	1.19	“难以”“ unable”“难于”“无法”“最难”“而无法”“ difficult”“困难的”
L20N3570-1 t0-v0.1	0.713	0.73	0.876	1.28	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.