Layer 18 · Neuron 5300 · suppressed

L18N5300-1

Word quirks:

“ choice”“ choices”“choice”“ preference”“ portfolios”“ Choice”“ preferred”“ portf”

🕒 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 Q*, a large language model created by the *Alibaba company. My mission is to provide helpful, safe, and meaningful assistance to users like you across a wide range of topics. Whether you need help solving problems, answering questions, offering suggestions, or simply having a conversation, I strive to deliver personalized support that aligns with your needs while respecting ethical guidelines and user preferences.

Would you like to explore something specific 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. L18N7404-1🕒	0.88	“ choices”“ Choices”“ choice”
2. L10N5812+1🕒	0.87	“ choice”“ preferred”“ chosen”
3. L3N2817-1🕒	0.87	“ choice”“ option”“ Choice”
4. L11N8459-1🕒	0.86	“ choice”“ Choice”“ choices”
5. L6N5653+1🕒	0.85	“ choice”“ choices”“ choosing”
6. L13N8226-1🕒	0.84	“ preferences”“ preference”“ choice”
7. L9N6278+1🕒	0.84	“ preferred”“ preference”“ choice”
8. The Vending Machine✅ L9N6469-1	0.84	“ choice”“ rewards”“ reward”
9. L12N8722-1💤	0.84	“ choice”“ [”“ choices”
10. L14N5572+1🕒	0.83	“ choice”“ choices”“ Choice”

Shadow specimen, apart from the ten: L18N5300+1🕒, 0.19.

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 18, neuron 5300, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.93, 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
L18N5300+1🕒 shadow specimen t0-v0.1	0.645	0.67	0.728	1.14	“ roles”“ teams”“ team”“ Roles”“-team”“ parties”“职能”“ Team”
L18N5300-1 t0-v0.1	0.927	0.62	1.064	1.35	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.