Layer 29 · Neuron 8874 · suppressed

L29N8874-1

Word quirks:

“包”“嘉”“ripe”“instrument”“委书记”“콘”“ihe”“zo”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am Qwen3.5, the latest large language model developed by Alibaba's team of researchers at Tao (通义). My core identity revolves around being a helpful, intelligent, and adaptable AI assistant designed to bridge human creativity with machine intelligence.

My Mission:

To empower users across diverse fields by:

Solving complex problems: From coding and data analysis to scientific research and creative writing.
Amplifying human potential: By automating routine tasks, providing real-time information, and offering personalized guidance.
**Promoting responsible

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. L29N8549+1🕒	0.75	“策”“社”“宗”
2. L29N6060+1💤	0.74	“富”“包”“ unf”
3. L28N3671+1💤	0.74	“希”“御”“沙”
4. L28N4146-1💤	0.74	“吹”“鼎”“青”
5. L26N769-1💤	0.74	“宗”“inser”“红”
6. L27N8750+1💤	0.74	“畜”“烟”“牛肉”
7. L27N6354+1🕒	0.74	“marker”“ostr”“ifero”
8. L30N1790+1💤	0.74	“营”“司”“宰”
9. L30N7387-1💤	0.74	“包”“包的”“etz”
10. L24N8575-1🕒	0.74	“不放”“豪”“坛”

Shadow specimen, apart from the ten: L29N8874+1🕒, 0.13.

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 8874, 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 0.81) 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
L29N8874+1🕒 shadow specimen t0-v0.1	0.408	0.78	0.455	1.51	“sess”“/pg”“ session”“/session”“ Session”“角色”“ngu”“engg”
L29N8874-1 t0-v0.1	0.334	0.81	0.427	1.63	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.