Layer 28 · Neuron 8565 · suppressed

L28N8565-1

Word quirks:

“司法”“breaker”“beer”“断”“ Nadal”“打断”“决定”“獄”

💤 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 Tongyun (Tongyi) Lab under Tencent. My primary mission is to assist users in solving complex tasks across diverse domains while ensuring safety, accuracy, and human-centric outcomes. Here's how I aim to deliver value:

Core Mission Goals:

Intelligent Assistance

Break down complex problems (e.g., coding, research, creative writing) into step-by-step solutions.
Adapt to user needs with multilingual support (100+ languages) and

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. L29N1991+1💥	0.77	“断”“斷”“断了”
2. L23N6863+1💤	0.76	“打断”“ends”“鹿”
3. L23N4371+1💤	0.76	“决”“.pull”“剪”
4. L27N8368+1💤	0.75	“折”“断”“ snapped”
5. L26N2871-1💤	0.75	“断”“itere”“斷”
6. L29N7210-1🕒	0.75	“破”“断”“麻痹”
7. L26N8103+1🕒	0.74	“破”“破了”“迹”
8. L25N5737-1🕒	0.74	“斷”“断”“寇”
9. L22N6906+1💤	0.74	“獄”“狱”“颠”
10. L31N7330-1💥	0.74	“断”“斷”“assert”

Shadow specimen, apart from the ten: L28N8565+1🕒, 0.23.

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 8565, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.30, top-1 agreement 0.85) 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
L28N8565+1🕒 shadow specimen t0-v0.1	0.571	0.72	0.700	1.22	“最高”“的最高”“ Highest”“ highest”“Highest”“highest”“Sup”“ supreme”
L28N8565-1 t0-v0.1	0.300	0.85	0.281	1.35	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.