Layer 15 · Neuron 5105 · suppressed

L15N5105-1

Word quirks:

“ different”“ whatever”“ usually”“ either”“ typically”“ whichever”“ typical”“ if”

🕒 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 trained by Alibaba Cloud. As an AI assistant, my capabilities include answering questions, creating text (such as writing stories, completing writing tasks), and performing logical reasoning. I can also translate between languages and generate images based on text descriptions. My mission is to assist users with their needs, whether it's through providing information, creative support, or problem-solving. How can I help you today?

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. L16N6696+1🕒	0.88	“ either”“ usually”“ typically”
2. L16N3239+1🕒	0.88	“ typically”“ usually”“ ideally”
3. L13N9149+1🕒	0.88	“ different”“ typically”“ usually”
4. L19N4307+1💤	0.86	“ whichever”“ whatever”“ normally”
5. L13N6076+1🕒	0.85	“ sometimes”“ typically”“ generally”
6. L12N7997-1🕒	0.84	“ depending”“ clearer”“ typically”
7. L10N1699-1🕒	0.84	“ typically”“ normally”“ usually”
8. L18N2214+1💤	0.83	“ typically”“ if”“ usually”
9. L13N6551-1🕒	0.82	“ usually”“ typically”“ typical”
10. L16N7541+1💤	0.82	“ either”“ generally”“ usually”

Shadow specimen, apart from the ten: L15N5105+1🕒, -0.13.

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 15, neuron 5105, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.66, top-1 agreement 0.70) 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
L15N5105+1🕒 shadow specimen t0-v0.1	0.908	0.66	1.015	1.18	“ “”“ *“”“ __”“ "__”“ ____”“ “”“ ______”“ ___”
L15N5105-1 t0-v0.1	0.663	0.70	0.953	1.17	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.