Layer 30 · Neuron 8306 · boosted

L30N8306+1

Word quirks:

“riot”“ away”“-out”“步”“ pasi”“ haus”“gart”“带去”

🕒 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 Qiani, a large-scale language model developed by Toutiao AI. My primary mission is to provide high-precision, safe, and useful assistance across a wide range of tasks, including:

Answering questions: From scientific concepts to everyday trivia, I strive for accuracy and clarity.
Creative writing: Crafting stories, poems, scripts, or professional documents with style and nuance.
Logical reasoning: Tackling math problems, analytical challenges, or step-by-step problem-solving.
Programming: Writing, debugging, and explaining code in multiple languages.

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. Downward Spiral✅ L18N1554-1	0.57	“下去”“soever”“仔”
2. L24N5673+1💤	0.54	“ pit”“值得”“ path”
3. L25N5482-1🕒	0.54	“债权债务”“azzi”“-out”
4. L26N818+1💤	0.54	“zas”“enberg”“打起”
5. L27N5932-1💤	0.53	“舷”“ Whip”“agerie”
6. L25N6828+1🕒	0.53	“-out”“ carriers”“源”
7. L29N2509+1💤	0.53	“ Purtroppo”“expédition”“羽”
8. L28N4311-1🕒	0.53	“ walk”“抽”“侧”
9. L19N7655-1💤	0.52	“ISK”“走出来”“ walked”
10. L30N4816-1💥	0.52	“走”“走的”“走了”

Shadow specimen, apart from the ten: L30N8306-1🕒, 0.14.

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 30, neuron 8306, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.44, top-1 agreement 0.74) 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
L30N8306+1 t0-v0.1	0.438	0.74	0.501	1.69	Top boosted tokens ↑
L30N8306-1🕒 shadow specimen t0-v0.1	0.449	0.81	0.524	1.85	“ij”“疾”“ijs”“ils”“rets”“uly”“zij”“urons”

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.