Layer 27 · Neuron 1388 · boosted

L27N1388+1

Word quirks:

“мато”“狱”“غلال”“łuż”“ rys”“稀”“珠”“atts”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I'm Q3, a large-scale ultra-accury (likely meant "ultra-accury" → ultra-accury is not a standard term; I suspect you might have meant ultra-accury as a typo for ultra-accurate or perhaps ultra-precise, but given the context of AI models, it's more likely you meant Q3 as a specific model name or there was a slight misphrasing). Let me clarify:

I am Q3, a large-scale AI model designed to assist with a wide range of 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. L30N4133-1💤	0.76	“زل”“宰”“涉”
2. L26N2181-1💤	0.75	“封建”“珠”“zung”
3. L28N6937-1💤	0.73	“庄”“底”“厉”
4. L26N4190-1💤	0.73	“ρασ”“uzion”“素”
5. L26N1768-1🕒	0.73	“狱”“nicz”“熟”
6. L28N5208+1💤	0.72	“鹿”“atsu”“轮”
7. L29N5480-1🕒	0.72	“葫芦”“pul”“扑”
8. L30N2373-1💤	0.72	“糊”“玉”“engt”
9. L22N3216-1💤	0.71	“ xlabel”“景”“税金”
10. L26N5652-1💤	0.71	“糟”“糊”“onio”

Shadow specimen, apart from the ten: L27N1388-1🕒, 0.17.

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 27, neuron 1388, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.40, 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
L27N1388+1 t0-v0.1	0.396	0.82	0.423	1.14	Top boosted tokens ↑
L27N1388-1🕒 shadow specimen t0-v0.1	1.710	0.46	1.823	1.79	“ objective”“客”“Objective”“ Objective”“objective”“客的”“ khách”“客观”

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.