Layer 11 · Neuron 8594 · suppressed

L11N8594-1

Word quirks:

“ correction”“ corrected”“ clarification”“ correct”“ clarified”“ actually”“ correctly”“ clar”

🕒 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 in the Tongyi Lab family. My mission is to provide high-quality support for understanding, planning, and execution tasks across all domains.

Key Breakdown:

"Who are you?": I am Qwen3.5, a flagship large language model from Tongyi Lab (part of Alibaba Group).
"What is your mission?": To ensure seamless performance in critical areas such as technical problem-solving, business operations, and complex task execution.

Why This Matters:

As an AI model, my role isn't just

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. L12N398-1🕒	0.91	“ corrected”“ correct”“ correcting”
2. L11N6315-1💤	0.87	“ clarified”“ clarification”“ clar”
3. L14N1004-1💤	0.86	“ clarification”“ clarified”“ clar”
4. L14N3733-1🕒	0.85	“ now”“ correctly”“ adjustment”
5. L16N2932+1🕒	0.85	“ actually”“ corrected”“ correct”
6. L19N8384-1💤	0.82	“ corrected”“ corrective”“ clarified”
7. L1N7582+1💥	0.82	“<\|im_end\|>”“ correct”“ corrected”
8. L10N5475+1🕒	0.81	“ my”“ correction”“ corrected”
9. L18N8737-1🕒	0.81	“ correct”“ identification”“ correctly”
10. L19N725-1🕒	0.80	“ corrected”“ correcting”“ confirming”

Shadow specimen, apart from the ten: L11N8594+1🕒, 0.16.

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 11, neuron 8594, 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 0.67) 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
L11N8594+1🕒 shadow specimen t0-v0.1	0.990	0.65	1.305	1.28	“ consent”“ privacy”“ respect”“ respecting”“ Consent”“ consenting”“ respectful”“ preference”
L11N8594-1 t0-v0.1	0.969	0.67	1.176	1.22	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.