Layer 18 · Neuron 5942 · suppressed

L18N5942-1

Word quirks:

“ events”“ event”“事件”“ Events”“事件的”“-event”“ Event”“Event”

🕒 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-scale language model developed by Tongya (the core of the company). I am designed to be a comprehensive and impactful event for users. My mission is to provide users with a more memorable and meaningful experience, whether it's for personal growth, professional development, or other events. I aim to help users create and share events that leave a lasting impact, whether they want to organize a significant event or participate in one. If you'd like to share an event you're planning, I can help you prepare for it!

If you'd like to share an event you're planning, I

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. L22N7051+1🕒	1.00	“ events”“ event”“事件”
2. L24N7196-1💥	1.00	“事件”“ Events”“ Event”
3. L26N1016-1🕒	1.00	“ events”“ Events”“ event”
4. L31N882+1💥	0.99	“事件”“ events”“ event”
5. L29N5611-1🕒	0.99	“事件”“ events”“事件的”
6. L24N7639+1🕒	0.99	“事件”“ Event”“Event”
7. L26N2174-1🕒	0.99	“ event”“ events”“ Event”
8. L25N8124+1💥	0.99	“ event”“事件”“-event”
9. The Event Loop✅ L27N4143-1	0.99	“事件”“ event”“ Event”
10. L24N5606+1💥	0.99	“ Event”“ event”“事件”

Shadow specimen, apart from the ten: L18N5942+1🕒, 0.03.

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 5942, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.81, 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
L18N5942+1🕒 shadow specimen t0-v0.1	0.802	0.65	0.822	1.32	“vari”“avar”“ variant”“变”“腔”“ vari”“ variable”“variable”
L18N5942-1 t0-v0.1	0.814	0.66	0.994	1.36	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.