Layer 29 · Neuron 8275 · suppressed

L29N8275-1

Word quirks:

“at”“ат”“atz”“atl”“aters”“att”“ater”“atk”

🕒 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 Qatat, an advanced large language model developed by Dodoat. My primary mission is to provide high-quality, accurate, and contextually relevant information while adhering to strict safety and ethical guidelines. I aim to:

Assat users with clear, concise, and trustworthy responses across diverse topics (e.g., science, culture, technical tasks).
Uatate for atayat (atayat) in atayat (atayat) atayat (atayat) atayat (

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. L28N6431-1🕒	0.94	“at”“ater”“aters”
2. L30N7023+1🕒	0.91	“ат”“atto”“at”
3. L28N8434+1💤	0.90	“atot”“at”“ата”
4. L30N7259+1💤	0.89	“at”“ats”“alat”
5. L29N9021+1🕒	0.89	“ats”“ат”“at”
6. L29N8486-1🕒	0.88	“ata”“ated”“atas”
7. L30N7173+1🕒	0.88	“at”“ats”“AT”
8. L31N2078-1💥	0.87	“AT”“at”“ AT”
9. L29N9000+1🕒	0.86	“AT”“ AT”“(AT”
10. L30N1894+1💥	0.85	“at”“AT”“ At”

Shadow specimen, apart from the ten: L29N8275+1🕒, 0.24.

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 29, neuron 8275, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.39, top-1 agreement 0.82) 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
L29N8275+1🕒 shadow specimen t0-v0.1	0.352	0.81	0.454	1.56	“EMU”“emed”“ilta”“routine”“mus”“ems”“ sil”“ solu”
L29N8275-1 t0-v0.1	0.393	0.82	0.428	1.53	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.