Layer 17 · Neuron 7650 · suppressed

L17N7650-1

Word quirks:

“ determines”“ determining”“ determine”“ defines”“ depends”“ Determines”“ determined”“ determination”

🕒 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, a large language model developed by Alibaba's AI technology company. My core capabilities focus on answering questions, performing logical reasoning, coding tasks, and handling multi-step problem-solving. I also support long-context understanding of documents and images, which helps in analyzing input data.

My mission is to assist users in completing tasks by providing accurate, relevant, and context-aware responses. This aligns with my design as a helpful assistant in the field of knowledge and problem-solving. How can I assist you with your specific needs?

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. L18N3226-1🕒	0.96	“ determines”“ determined”“ determine”
2. L5N1923+1🕒	0.94	“ determines”“ depends”“ determining”
3. L19N103+1🕒	0.94	“ determin”“ determining”“ determinant”
4. L16N7572-1🕒	0.94	“ determining”“ determination”“ deciding”
5. L15N213-1💤	0.93	“ determining”“ determined”“ determin”
6. L10N5011+1🕒	0.93	“ determining”“ decision”“ determines”
7. L6N2290+1🕒	0.93	“ factors”“ determines”“ factor”
8. L15N3616+1🕒	0.92	“ determining”“ determines”“ determination”
9. L12N8562+1🕒	0.92	“ determination”“ determining”“ clearer”
10. L16N2063+1💤	0.92	“ determines”“ determining”“ determination”

Shadow specimen, apart from the ten: L17N7650+1💤, 0.10.

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 17, neuron 7650, 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 0.67) 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
L17N7650+1💤 shadow specimen t0-v0.1	0.516	0.74	0.629	1.14	“ often”“⏎”“ sometimes”“ if”“ many”“ apparent”“ soon”“ both”
L17N7650-1 t0-v0.1	0.705	0.67	0.832	1.20	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.