Layer 9 · Neuron 6466 · boosted

L9N6466+1

Word quirks:

“ sometimes”“ usually”“ nowadays”“ basically”“ \"”“ question”“ somehow”“ always”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

Hello! I am Qwen3.5, a large language model developed by the Qwen team.

As for my mission, it's to provide users with high-quality, efficient, and helpful assistance. When you ask me "What are the advantages of Qwen3.5?", it means you're looking for a model that can offer more precise and accurate answers. Here are some specific advantages:

Enhanced Capabilities: I have been thoroughly trained on a comprehensive dataset, which has significantly improved my performance in understanding and generating accurate information.
Advanced Reasoning: Unlike other models, I can perform

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. L10N224+1🕒	0.86	“ basically”“ always”“ sometimes”
2. L12N4395+1💤	0.86	“ sometimes”“ actually”“ basically”
3. L11N465-1💤	0.85	“ sometimes”“ usually”“ basically”
4. L11N5588+1🕒	0.85	“ often”“ sometimes”“ generally”
5. L12N3358-1🕒	0.85	“ maybe”“ perhaps”“ often”
6. L11N7705+1🕒	0.85	“ somehow”“ sometimes”“ always”
7. L10N6262+1🕒	0.84	“ usually”“ typically”“ basically”
8. L13N3448+1💤	0.84	“ maybe”“ basically”“ sometimes”
9. L13N1700-1💤	0.84	“ "”“ usually”“ if”
10. L11N6763-1💤	0.84	“ usually”“ generally”“ sometimes”

Shadow specimen, apart from the ten: L9N6466-1🕒, 0.12.

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 9, neuron 6466, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L9N6466+1 t0-v0.1	1.250	0.57	1.540	1.47	Top boosted tokens ↑
L9N6466-1🕒 shadow specimen t0-v0.1	2.029	0.43	2.402	1.97	“ stories”“ story”“ narrative”“ narr”“ narratives”“ book”“ books”“ histories”

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.