Layer 30 · Neuron 8827 · boosted

L30N8827+1

Word quirks:

“锡”“ZE”“ze”“bage”“ذ”“�”“etra”“zes”

🕒 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. I am designed to assist users across a wide range of tasks, from creative writing and logical reasoning to code generation and data analysis.

Who am I?

Name: Qwen3.5 (Qianwen 3.5)
Developer: Alibaba Cloud's Tongyi Lab
Key Capabilities:
Advanced Reasoning: Solves complex math, physics, and logic problems.
Code Generation: Writes, debugs, and translates code in 10

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. L30N8977-1🕒	0.83	“ ze”“ Ze”“ze”
2. L28N3461+1💤	0.81	“ze”“zen”“z”
3. L31N3469+1💥	0.81	“za”“ze”“zu”
4. L29N8171+1🕒	0.80	“ ze”“ Ze”“Ze”
5. L29N8633-1🕒	0.80	“z”“ze”“zé”
6. L24N8022+1💤	0.80	“ZO”“ZE”“ ZE”
7. L27N6089+1💤	0.79	“ze”“zed”“zing”
8. L29N5806-1🕒	0.79	“z”“ete”“恩”
9. L29N8851+1🕒	0.78	“ZE”“zek”“Z”
10. L29N7308+1🕒	0.78	“苏”“zi”“zes”

Shadow specimen, apart from the ten: L30N8827-1🕒, 0.32.

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 8827, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.41, top-1 agreement 0.79) 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
L30N8827+1 t0-v0.1	0.410	0.79	0.435	1.83	Top boosted tokens ↑
L30N8827-1🕒 shadow specimen t0-v0.1	0.454	0.77	0.571	1.58	“단”“usher”“단의”“omi”“entai”“_cli”“omy”“OME”

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.