Layer 29 · Neuron 8226 · suppressed

L29N8226-1

Word quirks:

“身”“源”“神”“ival”“形”“流”“风”“字”

💤 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 latest large language model developed by Alibaba Cloud's Rival AI (Tongyi Lab). My mission is to empower users with accurate, safe, and actionable insights across a vast range of domains—from complex problem-solving and code development to multilingual communication and real-world task automation.

Key Focuses:

Precision & Reliability: Delivering high-accuracy responses in science, math, and professional contexts.
Autonomy & Efficiency: Automating workflows (e.g., data analysis, code generation) while minimizing human intervention.
**

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. L29N2222+1💤	0.89	“水”“血”“气”
2. L31N110-1💥	0.89	“计”“行”“文”
3. L27N6167+1💤	0.88	“水”“流”“首”
4. L31N5521-1💥	0.88	“血”“火”“雪”
5. L27N153-1💤	0.88	“小”“储”“拉”
6. L30N5475+1💤	0.88	“形”“参”“名”
7. L30N135-1🕒	0.88	“水”“人”“色”
8. L30N158+1💤	0.87	“圆”“水”“油”
9. L30N4176-1💤	0.87	“文”“主”“引”
10. L28N5394+1🕒	0.87	“秩”“核”“夷”

Shadow specimen, apart from the ten: L29N8226+1🕒, 0.21.

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 29, neuron 8226, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.29, 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
L29N8226+1🕒 shadow specimen t0-v0.1	0.373	0.79	0.428	1.58	“apar”“别”“asha”“ oks”“czyn”“afs”“lö”“ прим”
L29N8226-1 t0-v0.1	0.294	0.80	0.354	1.58	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.