Layer 28 · Neuron 9170 · suppressed

L28N9170-1

Word quirks:

“eness”“ogna”“en”“大臣”“uran”“ miss”“ato”“ATO”

💤 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, a large language model developed by TongYao (通义). My primary mission is to empower users with accurate, safe, and efficient assistance across a wide range of tasks—whether that involves solving complex problems, creating content, analyzing data, or simply having a conversation.

Key Missions:

Empower Human Potential:

I aim to act as a versatile tool for learning, creativity, and productivity, helping users overcome barriers like language limits, technical complexity, or time constraints.

Ensure Safety and Ethics:

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. L31N8889+1💥	0.67	“ико”“ATO”“orton”
2. L31N5617-1💥	0.67	“otton”“ugi”“enta”
3. L31N8438+1💥	0.66	“unan”“onat”“onale”
4. L29N7975+1🕒	0.66	“enan”“ارت”“enor”
5. L31N8782+1💥	0.66	“arro”“enson”“utre”
6. L31N5118-1💥	0.66	“onis”“ampp”“osta”
7. L31N8897-1💥	0.66	“ardi”“emos”“ozen”
8. L31N6407-1💥	0.64	“ух”“SION”“ammo”
9. L22N1528+1💤	0.64	“ proportion”“atos”“otec”
10. L31N9108-1💥	0.64	“уп”“utherford”“otka”

Shadow specimen, apart from the ten: L28N9170+1🕒, 0.15.

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 28, neuron 9170, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.26, top-1 agreement 0.80) 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
L28N9170+1🕒 shadow specimen t0-v0.1	1.028	0.62	1.124	1.28	“国家”“不同”“ National”“國家”“National”“国家政策”“年国家”“国家规定”
L28N9170-1 t0-v0.1	0.261	0.80	0.231	1.32	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.