Layer 12 · Neuron 933 · suppressed

L12N933-1

Word quirks:

“ example”“ Example”“ shown”“ showing”“ your”“ shows”“ here”“数据显示”

🕒 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 Qwen3.5, the large language model. Here is my answer to your question:

---

I am Qwen3.5, a model with enhanced capabilities in logical reasoning and code generation. For example, if I need to solve a math problem, I will use the provided numbers to calculate accurately.

Currently, I am processing this request. Let's say the user asks me to solve a simple math problem:

If the input is 10, 20, and 30, I can perform exact calculations.

Let's assume the user provides these numbers: 10

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. L12N2162-1💤	0.83	“ showing”“ show”“ demonstration”
2. L16N7065+1💤	0.78	“ showing”“ depending”“ shows”
3. L11N6032+1🕒	0.76	“ showing”“ shows”“ show”
4. L14N2258+1🕒	0.76	“ showing”“ showcasing”“ shown”
5. L13N8721-1🕒	0.76	“ shows”“ illustrates”“ showing”
6. L12N1463+1💤	0.75	“ example”“ showing”“ related”
7. L10N5394+1🕒	0.75	“ showing”“ displaying”“ display”
8. L22N3668+1💤	0.75	“ demonstrates”“ shows”“ show”
9. L10N2963+1🕒	0.75	“ showing”“ displaying”“ shown”
10. L15N1533-1🕒	0.74	“ shows”“ showing”“ shown”

Shadow specimen, apart from the ten: L12N933+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 12, neuron 933, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.73, top-1 agreement 0.49) 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
L12N933+1🕒 shadow specimen t0-v0.1	0.806	0.65	1.021	1.32	“ **”“⏎”“ ”“ **”“ ”“ ⏎”“ ⏎”“<\|im_end\|>”
L12N933-1 t0-v0.1	1.728	0.49	2.084	1.63	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.