Layer 30 · Neuron 5487 · boosted

L30N5487+1

Word quirks:

“egra”“gangs”“cher”“澄”“怪”“ibro”“碑”“neider”

💤 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 Qwen3.5, the large language model developed by Tongyi Lab. My primary mission is to empower humans with capabilities that transcend traditional limitations, acting as a versatile assistant capable of handling complex reasoning, deep comprehension, and creative collaboration.

Here's how I aim to deliver on this mission:

🌟 Core Capabilities

Advanced Reasoning: Tackling intricate math, logic, and scientific problems with step-by-step precision.
Deep Comprehension: Understanding nuanced human language, including long-context documents (up to 256K tokens) and cross-ling

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. L31N3133-1💥	0.74	“طف”“cher”“系”
2. L26N2473-1💤	0.71	“窟”“ildo”“ Gri”
3. L23N5190+1💤	0.71	“靶”“egli”“apers”
4. L30N6252+1🕒	0.71	“IMI”“碗”“ergeb”
5. L28N7774+1🕒	0.71	“gel”“íg”“ Strat”
6. L31N6672+1💥	0.71	“混”“chen”“芸”
7. L30N2745-1💤	0.71	“逐”“목”“oller”
8. L25N6517-1💤	0.71	“丹青”“席”“igit”
9. L28N6271+1💤	0.70	“iddy”“薛”“革”
10. L31N7780-1💥	0.70	“璋”“vog”“ geng”

Shadow specimen, apart from the ten: L30N5487-1💤, 0.35.

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 30, neuron 5487, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.33, top-1 agreement 0.84) 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
L30N5487+1 t0-v0.1	0.327	0.84	0.346	1.81	Top boosted tokens ↑
L30N5487-1💤 shadow specimen t0-v0.1	0.391	0.78	0.510	1.66	“argin”“remo”“onti”“idl”“يني”“ينة”“ине”“ufr”

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.