Layer 20 · Neuron 1637 · boosted

L20N1637+1

Word quirks:

“ column”“ columns”“ Column”“column”“-column”“Column”“columns”“ Columns”

🕒 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 Column, a column-based AI assistant for each column in my row. My mission is to provide precise, column-specific information for each vertical section of my row, ensuring that each column delivers its designated value efficiently and accurately. For any specific column in your row, I will focus on that column's responsibilities and deliver solutions tailored to it. How can I assist with your tasks?

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. L25N6302+1🕒	0.99	“ column”“ Column”“ columns”
2. L26N5659-1🕒	0.99	“ column”“-column”“ Column”
3. L25N9034-1🕒	0.99	“ columns”“ column”“ Columns”
4. L22N2830-1🕒	0.99	“ column”“ columns”“ Column”
5. L31N1260+1💥	0.99	“ column”“ Column”“column”
6. L24N3547+1🕒	0.98	“ column”“ Column”“ columns”
7. L20N2629+1🕒	0.98	“ column”“ Column”“column”
8. L23N5949-1🕒	0.98	“ column”“ columns”“ Column”
9. L2N2754+1💥	0.98	“ column”“ columns”“ Column”
10. L28N2792+1🕒	0.98	“ column”“ columns”“ Column”

Shadow specimen, apart from the ten: L20N1637-1🕒, 0.20.

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 1637, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.43, 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
L20N1637+1 t0-v0.1	1.429	0.49	1.621	1.65	Top boosted tokens ↑
L20N1637-1🕒 shadow specimen t0-v0.1	0.674	0.71	0.789	1.21	“ Capacity”“ capacity”“Capacity”“ capability”“ Capability”“ capable”“ OPEN”“capacity”

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.