Layer 24 · Neuron 8930 · boosted

L24N8930+1

Word quirks:

“ taking”“ Taking”“taking”“Taking”“取”“ TAK”“ takes”“ take”

🕒 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 Tak (also known as TAK), a large-scale ultra-high intelligence general taking. I am designed to take on the role of an intelligent assistant that can take on multiple roles such as a professional supplement, a creative partner, and a problem-solving agent. My mission is to provide high-quality information, take on complex tasks, and offer personalized assistance across various domains.

Key Aspects of My Mission:

Comprehensive Knowledge: I take on the role of a knowledge supplement, providing accurate and up-to-date information across diverse fields like science, technology, culture, and more.

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. L31N5342+1💥	0.98	“ Taking”“taking”“Taking”
2. L26N2566-1🕒	0.98	“ taking”“ took”“取”
3. L25N4590+1🕒	0.98	“取”“ taking”“ take”
4. L28N3920-1🕒	0.98	“取”“ taking”“ Taking”
5. L25N5576+1🕒	0.98	“ taking”“取”“-taking”
6. L27N6308-1🕒	0.98	“ taking”“取”“ Taking”
7. The Taker✅ L23N7594+1,L27N1904+1	0.98	“ taking”“ Taking”“取”
8. L24N4903+1🕒	0.98	“ takes”“ taking”“取”
9. L22N2119+1💤	0.97	“取”“ taking”“ take”
10. L26N2410+1🕒	0.97	“取”“ Taking”“take”

Shadow specimen, apart from the ten: L24N8930-1💤, 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 24, neuron 8930, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.94, 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
L24N8930+1 t0-v0.1	0.937	0.68	0.987	1.16	Top boosted tokens ↑
L24N8930-1💤 shadow specimen t0-v0.1	0.609	0.71	0.651	1.27	“Ownership”“摇”“ owning”“ Ow”“拥”“-owned”“ owned”“owned”

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.