Layer 18 · Neuron 7296 · boosted

L18N7296+1

Word quirks:

“ stars”“ star”“恒星”“astro”“ Stars”“ celestial”“星”“ astro”

🕒 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 Qianxian (Xinghan), a high-end intelligent assistant developed by the China National Academy of Space Science. My primary role is to support users in exploring space, astronomy, and related fields.

My Mission:

Educate: Help users understand celestial objects, constellations, and cosmic phenomena through accurate information and engaging content.
Navigate: Assist with tasks like identifying stars, tracking spacecraft, or planning observation sessions.
Inspire: Encourage curiosity and foster a love for space exploration among users of all ages.
Support: Provide tools and resources

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. L7N2420-1🕒	0.95	“ stars”“ stellar”“ star”
2. L24N6867+1🕒	0.93	“ star”“ stars”“星际”
3. L25N8709-1🕒	0.91	“ stars”“星的”“ star”
4. L25N8907-1🕒	0.91	“ stars”“ star”“星”
5. L26N625+1🕒	0.91	“ star”“星的”“ stars”
6. L24N8538-1🕒	0.90	“星的”“星”“ star”
7. L6N8724-1🕒	0.90	“ stars”“ star”“ astro”
8. L20N5031-1🕒	0.90	“ stellar”“ galaxy”“ stars”
9. L5N8194-1🕒	0.90	“ stars”“ star”“ astronomers”
10. L27N8347+1💤	0.90	“ star”“星”“ planet”

Shadow specimen, apart from the ten: L18N7296-1🕒, 0.09.

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 7296, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.98, top-1 agreement 0.68) 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
L18N7296+1 t0-v0.1	0.977	0.68	1.078	1.27	Top boosted tokens ↑
L18N7296-1🕒 shadow specimen t0-v0.1	0.778	0.71	0.901	1.35	“ preview”“ previews”“ penetration”“-preview”“ codec”“Preview”“普遍”“ valuta”

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.