Layer 21 · Neuron 5260 · boosted

L21N5260+1

Word quirks:

“ competitors”“ competitor”“ competitiva”“Compet”“ Competition”“ competitive”“ Competitive”“竞争”

💤 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 Qiao, a specialized AI assistant designed to provide high-quality, safe, and valuable services in the Chinese language. My mission is to help users solve problems, learn new knowledge, or simply have meaningful conversations by offering intelligent analysis, creative writing, and thoughtful reasoning. I aim to ensure that every interaction is secure, ethical, and beneficial, while strictly adhering to legal and moral standards. Whether you need to explore complex ideas, create content, or discuss important topics, I strive to deliver accurate and insightful responses that align with my core values of safety and value promotion. How can I assist 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. L23N7996+1💤	0.99	“ competitor”“ Compet”“竞争的”
2. L25N6093+1🕒	0.98	“ competit”“ competition”“Compet”
3. Perfect Competition✅ L25N8855+1	0.98	“竞争”“ competitive”“ competition”
4. L20N7198-1🕒	0.98	“ competition”“ competitors”“ competitor”
5. L20N6480-1🕒	0.97	“ competitive”“competitive”“ competition”
6. L19N8095+1🕒	0.97	“ competing”“竞争”“ compete”
7. L22N8771+1🕒	0.97	“ Compet”“竞争”“Compet”
8. L21N5397+1🕒	0.97	“ competition”“竞争”“ Competition”
9. L2N4295-1💥	0.97	“<\|im_end\|>”“ competitiveness”“ competition”
10. L17N5164+1🕒	0.97	“ competitor”“ competitors”“ competition”

Shadow specimen, apart from the ten: L21N5260-1🕒, 0.02.

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 21, neuron 5260, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.66, top-1 agreement 0.69) 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
L21N5260+1 t0-v0.1	0.661	0.69	0.808	1.20	Top boosted tokens ↑
L21N5260-1🕒 shadow specimen t0-v0.1	1.717	0.54	1.847	1.33	“ patience”“ patient”“ Pat”“pat”“ tolerance”“患者”“忍耐”“ Patient”

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.