Layer 12 · Neuron 5625 · suppressed

L12N5625-1

Word quirks:

“ describing”“ abstract”“ description”“ imagined”“ describe”“ imaginary”“ unusual”“ imagining”

🕒 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 latest large language model developed by Alibaba Cloud's Tongyi Lab. My mission is to demonstrate my capabilities as a cutting-edge AI assistant while highlighting my unique strengths. I aim to showcase my ability to understand complex instructions and generate creative content that reflects my advanced training and deep understanding of the world.

My goal is to provide a seamless user experience that feels natural and intuitive, making me accessible to anyone who can articulate their needs in a clear and concise manner. By focusing on my core competencies, I strive to create a memorable impression that captures the essence of my identity as an AI model designed

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. L15N6678+1🕒	0.88	“ describing”“ imagined”“ imaginary”
2. L12N5368+1💤	0.88	“ describing”“ description”“ descriptions”
3. L11N354-1💤	0.87	“ imagined”“ hypothetical”“ imaginary”
4. L8N2747+1🕒	0.87	“ imagined”“ description”“ __”
5. L17N7118-1💤	0.85	“ description”“ imagined”“ fictional”
6. L17N7955+1🕒	0.84	“ description”“ describing”“想象”
7. L10N1046+1🕒	0.84	“ ideal”“ because”“ describing”
8. L12N1837-1💤	0.83	“ mentioned”“ imagined”“ described”
9. L8N1476-1🕒	0.82	“ metaphor”“ description”“ imagining”
10. L12N8387+1🕒	0.80	“ often”“ textual”“ detailed”

Shadow specimen, apart from the ten: L12N5625+1🕒, 0.17.

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 5625, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.87, 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
L12N5625+1🕒 shadow specimen t0-v0.1	1.529	0.54	1.765	1.31	“ referenced”“ mentioned”“ accompanying”“ available”“ downloadable”“ supporting”“ corresponding”“ provided”
L12N5625-1 t0-v0.1	0.870	0.70	1.105	1.27	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.