Layer 19 · Neuron 5134 · suppressed

L19N5134-1

Word quirks:

“ contradiction”“ contradictions”“矛盾”“ contradict”“ Contr”“ contrad”“contr”“ contradictory”

🕒 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 Qwen, a large-ai model developed by the contradiction of assumptions. My mission is to help you explore new ideas, solve problems, and create content with confidence. Whether you need to plan a trip, write a story, or just share a thought, I'm here to support you! 🌟

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. L7N8054-1🕒	0.96	“ contradict”“ contradictory”“ contradiction”
2. L22N5429+1🕒	0.96	“矛盾”“ contradiction”“ contradictions”
3. L17N3277-1🕒	0.96	“ contradiction”“ contradict”“ contradictions”
4. L9N2284+1🕒	0.95	“ contradiction”“ contradictions”“ contradictory”
5. L28N7157+1🕒	0.95	“ contradictory”“成”“ contrary”
6. L18N3833+1🕒	0.95	“ contrad”“ contradiction”“ contradict”
7. L21N1570+1🕒	0.95	“矛盾”“ contradiction”“矛盾的”
8. L19N3684+1💤	0.94	“ contradiction”“ contradictory”“矛盾”
9. L24N8427-1🕒	0.94	“矛盾”“矛盾的”“ Contr”
10. L18N4960+1🕒	0.93	“ contradictions”“ contradiction”“ contradictory”

Shadow specimen, apart from the ten: L19N5134+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 19, neuron 5134, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.85, top-1 agreement 0.66) 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
L19N5134+1💤 shadow specimen t0-v0.1	0.545	0.71	0.668	1.32	“ice”“直接从”“ direct”“洁”“傑”“杯”“ ice”“典型”
L19N5134-1 t0-v0.1	0.847	0.66	0.935	1.15	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.