Layer 23 · Neuron 4687 · boosted

L23N4687+1

Word quirks:

“ vio”“vi”“ubu”“高压”“ nid”“筱”“分身”“ viol”

💤 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 Qwen3, a high-quality multi-vision and high-vocabulary countercapsule model developed by Alibaba's AI division. My main function is to help users interact with AI in a more intelligent and efficient way.

My mission is to:

Enhance user experience: Provide clear, accurate, and helpful responses across diverse fields like language translation, code generation, and data analysis.
Support real-world applications: Work with users to solve practical problems, from writing code to planning travel itineraries.
Maintain high standards: Deliver consistent performance with strong memory capabilities

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. The Salesman✅ L21N748-1	0.69	“ Vi”“kur”“vi”
2. L27N2763+1💤	0.68	“ V”“vodu”“XA”
3. L28N3822-1💤	0.67	“农作物”“ inve”“精灵”
4. L26N8099-1💤	0.67	“V”“威”“房”
5. L31N8686-1💥	0.66	“媛媛”“vro”“麒”
6. L29N7988+1🕒	0.66	“ vi”“ Vi”“vi”
7. L29N8536+1🕒	0.65	“ vi”“ Vi”“vi”
8. L29N8782-1🕒	0.65	“angen”“不死”“cr”
9. L30N8799+1💤	0.65	“VO”“伏”“صال”
10. L22N6475+1🕒	0.65	“ vind”“прово”“ ty”

Shadow specimen, apart from the ten: L23N4687-1🕒, 0.37.

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 23, neuron 4687, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.71, top-1 agreement 0.77) 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
L23N4687+1 t0-v0.1	0.708	0.77	0.761	1.29	Top boosted tokens ↑
L23N4687-1🕒 shadow specimen t0-v0.1	0.810	0.70	0.889	1.29	“erat”“ Embedded”“准”“ogar”“要跟”“ Embed”“推断”“ story”

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.