Layer 23 · Neuron 8048 · suppressed

L23N8048-1

Word quirks:

“ leap”“ Leap”“Leap”“ leaps”“-le”“_le”“.le”“le”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am leap2, a powerful business leap engine developed by Leap. My mission is to help businesses navigate the complexities of modern commerce by providing intelligent, actionable insights and streamlined operations. I focus on:

Business Optimization: Analyzing data to identify opportunities for leapfrogging in market positioning, cost reduction, and revenue enhancement.
Decision Support: Providing non-leap, non-jump, non-leap, non-jump, non-leap, non-jump, non-leap, non-jump, non-leap, non-jump, non-leap, non-jump,

The AI judge wasn't sure this holds together.

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. L25N619-1🕒	0.80	“ le”“ Le”“Le”
2. L28N5422+1🕒	0.79	“ Le”“ le”“Le”
3. L28N1883-1🕒	0.79	“Le”“le”“ Le”
4. L29N7466+1🕒	0.79	“le”“-le”“LE”
5. L28N2531-1💤	0.79	“Le”“ LE”“ Le”
6. L29N7092-1🕒	0.79	“ le”“-le”“le”
7. L28N8470-1🕒	0.79	“ Le”“ le”“Le”
8. L29N8740+1🕒	0.78	“ le”“le”“ Le”
9. L29N7003-1🕒	0.78	“Le”“ le”“ Le”
10. L26N2662+1🕒	0.78	“ le”“ Le”“ len”

Shadow specimen, apart from the ten: L23N8048+1🕒, 0.16.

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 23, neuron 8048, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L23N8048+1🕒 shadow specimen t0-v0.1	0.836	0.68	0.923	1.23	“ HOME”“HOME”“roje”“前去”“ bd”“allah”“ housing”“home”
L23N8048-1 t0-v0.1	1.197	0.62	1.254	1.12	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.