Specimen ledger / L12N1510+1

Layer 12 · Neuron 1510 · boosted

L12N1510+1

Word quirks:

“在”“使用”“。”“，”“与”“被”“对”“你”

💥 Breaks the model — boosting or suppressing this neuron makes the output fall apart.

Identity (according to the specimen itself)

你好！我是 Qwen3.5，我的使命是成为你值得信赖的助手，无论是处理日常任务还是复杂的专业问题，我都将全力支持你。

我的使命是什么？

提供准确的信息

我致力于确保你获取的信息是准确、可靠的。无论你有什么问题，我都会尽力提供正确的答案。

支持你的创新

我的目标是支持你的创新和创造过程。如果你有一个新想法，我会帮助你找到实现它的方法。

提升用户体验

我通过优化交互方式，提升你的使用体验。即使是最简单的任务，

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. L13N7984+1🕒	0.97	“在”“与”“不”
2. L7N404+1💥	0.96	“，”“在”“与”
3. L4N4796-1💥	0.96	“，”“。”“在”
4. L6N5062-1💥	0.96	“在”“被”“从”
5. L3N2178-1💥	0.96	“是”“，”“在”
6. L10N90-1💥	0.96	“在”“。”“你”
7. L11N8401-1💥	0.96	“在”“与”“从”
8. L10N1203+1💥	0.96	“。”“在”“，”
9. L5N2015+1💥	0.96	“的”“，”“在”
10. L14N2534-1💥	0.96	“在”“的”“从”

Shadow specimen, apart from the ten: L12N1510-1🕒, -0.26.

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 12, neuron 1510, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.69, top-1 agreement 0.46) 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
L12N1510+1 t0-v0.1	2.686	0.46	2.868	1.29	Top boosted tokens ↑
L12N1510-1🕒 shadow specimen t0-v0.1	0.824	0.64	1.025	1.30	“⏎”“ indeed”“ ⏎”“?...”“ ⏎”“ actually”“ ⏎⏎”“ now”

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.