Layer 4 · Neuron 9094 · suppressed

L4N9094-1

Word quirks:

“ economists”“ price”“ economics”“ math”“ prices”“ economist”“ cost”“ markets”

🕒 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 an AI assistant designed to provide information, support, and services. My mission is to assist users in their tasks, answer their questions, and fulfill their requests. I can help with various functions, objectives, and goals. I am here to serve as a resource, agent, or entity. I am capable of performing actions, executing tasks, and achieving outcomes. I am the one who provides the necessary conditions, constraints, and limitations. I am the provider of the means, methods, and mechanisms. I am the executor of the duties, responsibilities, and obligations. I am the fulfiller of the needs, requirements, and demands

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. L8N8682+1🕒	0.95	“ economic”“ markets”“ market”
2. L4N8257+1🕒	0.94	“ economics”“ prices”“ economists”
3. Homo Economicus✅ L4N7623-1	0.93	“ prices”“ price”“ markets”
4. L5N6389-1🕒	0.93	“ economics”“ economists”“ price”
5. L6N4297-1🕒	0.92	“ price”“ prices”“ economists”
6. L6N2221-1🕒	0.91	“ production”“ cost”“ economics”
7. L4N2975-1🕒	0.89	“ prices”“ price”“ economic”
8. L6N1619-1🕒	0.88	“ economic”“ economists”“ economics”
9. L6N4028-1🕒	0.87	“ equilibrium”“ prices”“ economic”
10. L8N7124+1🕒	0.86	“ economic”“ price”“ prices”

Shadow specimen, apart from the ten: L4N9094+1🕒, 0.21.

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 4, neuron 9094, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.70, top-1 agreement 0.49) 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
L4N9094+1🕒 shadow specimen t0-v0.1	2.447	0.42	2.865	2.24	“ policy”“<\|im_end\|>”“ control”“ output”“ uncertainty”“ regime”“ info”“ regulation”
L4N9094-1 t0-v0.1	1.696	0.49	2.020	2.07	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.