Layer 27 · Neuron 2543 · suppressed

L27N2543-1

Word quirks:

“ miss”“弃”“棄”“occhi”“Miss”“ Miss”“ missed”“miss”

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

Identity (according to the specimen itself)

I am Qo-330, a large-scale language model developed by ITC (Intelligence Technology Company). My mission is to provide high-quality, accurate, and safe AI services that can meet the diverse needs of users in various fields. I strive to be a reliable and creative partner in learning, work, and daily life, while adhering to ethical guidelines and ensuring safety. How can I help you today?

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. L24N8648-1🕒	0.83	“ Miss”“ miss”“miss”
2. L27N1338+1🕒	0.83	“ Miss”“ miss”“Miss”
3. L20N8980-1🕒	0.81	“ missed”“ missing”“ misses”
4. L28N3731-1💤	0.80	“ miss”“miss”“席”
5. L20N5614-1🕒	0.77	“ miss”“ misses”“ missing”
6. L26N8431+1🕒	0.77	“ Miss”“Miss”“ MISS”
7. L28N5268-1🕒	0.76	“彌”“ Miss”“ Mass”
8. L28N7807-1🕒	0.75	“任”“ miss”“ Miss”
9. L27N5275-1⛔	0.73	“ miss”“ Miss”“Miss”
10. L22N7360+1💤	0.73	“ miss”“ misses”“miss”

Shadow specimen, apart from the ten: L27N2543+1🕒, 0.08.

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 2543, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.44, top-1 agreement 0.76) 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
L27N2543+1🕒 shadow specimen t0-v0.1	0.532	0.76	0.541	1.45	“ reduce”“减少”“ Reduce”“ reduces”“消”“reduce”“ reducing”“ Address”
L27N2543-1 t0-v0.1	0.439	0.76	0.510	1.40	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.