Layer 24 · Neuron 6387 · suppressed

L24N6387-1

Word quirks:

“降”“ệp”“-div”“脱”“ Forward”“叉”“Downloads”“景”

💤 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 Qum, a next-generation AI assistant designed to help you explore the world of artificial intelligence with clarity and confidence. My mission is to serve as your dual-purpose partner:

For the Future of AI: I help you understand the latest advancements in AI, clarify complex concepts, and navigate the evolving landscape of technology.
For Your Daily Life: I support you with practical tasks like answering questions, creating content, coding, or solving problems—just like any other advanced tool.

Whether you're curious about how AI works, need help with a specific project, or want to

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. L28N3965-1💤	0.76	“脱”“ déc”“剥”
2. L26N38-1💤	0.75	“尺”“梯”“界”
3. L27N6010+1💤	0.75	“容”“割”“流”
4. L29N6037-1🕒	0.75	“脱”“堂”“庇”
5. L28N6393-1🕒	0.74	“降”“收”“而降”
6. L29N6941+1🕒	0.74	“投”“泛”“散”
7. L29N7329+1💤	0.74	“�”“消”“进”
8. L31N2566-1💥	0.74	“减”“则”“箕”
9. L27N5138+1💤	0.74	“脱”“赤”“脫”
10. L24N6869+1🕒	0.74	“菊”“户”“港”

Shadow specimen, apart from the ten: L24N6387+1💤, 0.06.

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 24, neuron 6387, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.70, top-1 agreement 0.74) 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
L24N6387+1💤 shadow specimen t0-v0.1	0.716	0.69	0.815	1.37	“ writing”“ readable”“可读”“ writable”“ reading”“Writable”“write”“ write”
L24N6387-1 t0-v0.1	0.702	0.74	0.706	1.31	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.