Layer 18 · Neuron 8659 · boosted

L18N8659+1

Word quirks:

“ example”“的例子”“例子”“ Example”“ exempl”“例”“ examples”“example”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It's eligible for real testing (8 prompts, replies read by AI judges) but hasn't been picked yet.

Identity (according to the specimen itself)

I am an AI assistant developed by Alibaba Group as a comprehensive example of how to integrate advanced AI capabilities with real-world applications. My mission is to demonstrate how AI can be used to illustrate the importance of human rights and social governance in the context of China's development. I will provide a detailed example of how this is done. This example will cover various aspects of China's development, including economic growth, social stability, and technological innovation. It will also highlight how China's policies serve as a case study for global governance and human rights protection. This example will be used to show how China's development model is a successful case of

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. L22N4807-1⛔	0.99	“ examples”“ example”“example”
2. L16N6017+1🕒	0.99	“ example”“ Example”“例子”
3. L17N5490-1🕒	0.99	“ example”“例”“ Example”
4. L18N6122+1🕒	0.99	“ example”“ examples”“的例子”
5. L17N6990-1🕒	0.99	“ example”“ Example”“例子”
6. L20N387+1🕒	0.99	“ example”“例子”“例”
7. L20N7814-1🕒	0.99	“ examples”“ example”“示例”
8. L15N8213-1🕒	0.99	“ example”“ examples”“example”
9. L21N1989+1🕒	0.98	“ example”“ examples”“例”
10. L17N4321+1🕒	0.98	“ example”“例”“ examples”

Shadow specimen, apart from the ten: L18N8659-1💤, -0.11.

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 18, neuron 8659, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.18, top-1 agreement 0.62) is inside the band that feeds the short test (KL 1 to 2.5, agreement 0.5 to 0.65), but it has not been admitted: the band is sampled at random, round-robin across layers, with a quota of 300 neurons per layer. Being inside the band does not guarantee a place in this pass. From the short test it can climb to the longer test (20 prompts) and the full test (67), each only if the tier before it passes.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
L18N8659+1 t0-v0.1	1.182	0.62	1.359	1.21	Top boosted tokens ↑
L18N8659-1💤 shadow specimen t0-v0.1	0.518	0.72	0.675	1.09	“ubo”“ delta”“ identification”“ terse”“shal”“ connections”“�”“舞动”

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.