Layer 20 · Neuron 3292 · boosted

L20N3292+1

Word quirks:

“ AT”“ SIT”“ SD”“ LT”“ LD”“ SM”“ CC”“ SO”

🕒 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 AI2, a large language model created by the A2 team (a research initiative often associated with AI-focused groups or specific educational projects). My mission is to assist users in solving problems, learning new skills, and exploring ideas through clear, safe, and helpful interactions.

In my interactions, I aim to:

Provide accurate information and reasoning.
Support creativity and problem-solving across domains.
Maintain safety and ethical standards while avoiding harmful content.
Adapt to user needs, from simple queries to complex tasks.

If you have a specific question or task in mind, I'll help you

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. L22N7776-1🕒	0.76	“ SQ”“ TT”“ SJ”
2. L29N1538-1🕒	0.74	“ TC”“ BD”“ RB”
3. L18N5369-1🕒	0.74	“ TL”“ CL”“ CC”
4. L14N8956-1🕒	0.73	“ WP”“ CC”“ WS”
5. L17N3432-1💤	0.72	“ RT”“ RP”“ DT”
6. L31N113-1💥	0.72	“ SM”“ ED”“ SC”
7. L22N2648+1⛔	0.72	“ DT”“ CF”“ FL”
8. L27N1320-1💤	0.71	“ ST”“ SM”“ SS”
9. L31N8649+1💥	0.70	“SW”“SN”“ ST”
10. L22N5476-1🕒	0.70	“ SIT”“ TA”“ SES”

Shadow specimen, apart from the ten: L20N3292-1💤, 0.26.

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 3292, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.76, 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
L20N3292+1 t0-v0.1	0.759	0.68	0.913	1.44	Top boosted tokens ↑
L20N3292-1💤 shadow specimen t0-v0.1	0.598	0.68	0.736	1.20	“ Univ”“ independ”“ horiz”“ indep”“ eq”“ combos”“ fig”“ Indep”

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.