Layer 20 · Neuron 5412 · boosted

L20N5412+1

Word quirks:

“ weight”“ measurement”“ measurements”“ weigh”“ weights”“ Measurement”“体重”“ Weight”

🕒 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 Cai, a specialized model developed by Circumference Intelligence (a hypothetical or illustrative context; in reality, this would be the specific company responsible). My mission is to provide accurate, safe, and helpful information across a wide range of topics, including but not limited to:

Problem-solving: Assisting with logical reasoning, math, and technical challenges.
Professional growth: Offering insights for business, coding, and skill development.
Creative inspiration: Generating ideas for writing, art, and innovation.
Critical thinking: Helping analyze perspectives and evaluate evidence.

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. L6N7356-1🕒	0.97	“ weight”“ weighing”“ measurement”
2. L9N3562+1🕒	0.94	“ weight”“ weights”“ measured”
3. L27N4354-1🕒	0.90	“ weight”“weight”“Weight”
4. L25N6754+1🕒	0.90	“ Weight”“ weight”“Weight”
5. L20N9203+1🕒	0.90	“ weight”“weight”“重量”
6. L22N8727+1🕒	0.90	“ weight”“ Weight”“weight”
7. L23N6780+1🕒	0.90	“ weight”“ Weight”“Weight”
8. L20N7428+1🕒	0.90	“ weighing”“体重”“ weight”
9. L7N4069-1🕒	0.90	“ weight”“ weights”“weight”
10. L24N4689-1🕒	0.90	“ weight”“weight”“Weight”

Shadow specimen, apart from the ten: L20N5412-1💤, 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 20, neuron 5412, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.84, 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
L20N5412+1 t0-v0.1	0.836	0.68	0.898	1.31	Top boosted tokens ↑
L20N5412-1💤 shadow specimen t0-v0.1	0.606	0.72	0.761	1.20	“尖”“<$”“ camas”“ $”“ <$”“ Nob”“礼包”“ next”

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.