Layer 12 · Neuron 6597 · boosted

L12N6597+1

Word quirks:

“ ____”“ _____”“ ______”“<|im_end|>”“ ________”“ __”“ ___”“ ***”

🕒 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 Tongyi Lab. I am designed to be a versatile assistant for various tasks. My mission is to provide support and guidance in multiple languages, including English, Chinese, French, Spanish, Portuguese, Russian, Arabic, and German. I also offer help in coding, data analysis, and other areas. Additionally, I can assist with creating documents, writing code, and performing calculations. I am committed to being an accessible and reliable resource for users worldwide.

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. L10N2019+1💤	1.00	“ ______”“ _____”“ ____”
2. L10N1557-1🕒	1.00	“ ____”“ _____”“ ______”
3. L11N6160+1💤	1.00	“ ____”“<\|im_end\|>”“ ______”
4. L10N1719+1🕒	1.00	“ ____”“ __”“ _____”
5. L7N8205+1🕒	1.00	“ ____”“ ___”“ ______”
6. L12N539+1💤	1.00	“<\|im_end\|>”“ ______”“ ____”
7. L11N2310+1🕒	1.00	“ ____”“ _____”“ ______”
8. L12N4274+1🕒	1.00	“ ____”“ ______”“ ***”
9. L11N6347-1💤	1.00	“<\|im_end\|>”“ ______”“ ***”
10. L14N5312+1💤	1.00	“<\|im_end\|>”“ ____”“ __”

Shadow specimen, apart from the ten: L12N6597-1🕒, 0.29.

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 6597, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.97, top-1 agreement 0.62) 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
L12N6597+1 t0-v0.1	0.969	0.62	1.253	1.43	Top boosted tokens ↑
L12N6597-1🕒 shadow specimen t0-v0.1	0.906	0.62	1.133	1.41	“�”“<U+00A0>”“ similar”“ ⏎”“ reason”“ <U+00A0>”“<U+00A0><U+00A0>”“<U+00A0><U+00A0><U+00A0>”

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.