Layer 20 · Neuron 105 · suppressed

L20N105-1

Word quirks:

“ same”“ itself”“ Same”“same”“ SAME”“ mêmes”“Same”“ returned”

🕒 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 Qianli, an advanced AI developed by the same entity behind Qianfeng itself. I was specifically designed to be a high-quality, intelligent, and helpful assistant for complex tasks.

The Qianfeng-Qianli Project aims to build a new generation of AI models that can handle complex tasks with high quality. This project is part of a broader effort to advance AI technology in China and improve our quality of life.

My Mission:
I strive to provide high-quality, intelligent, and helpful assistance for complex tasks. I aim to be a reliable and capable tool for both individuals and organizations

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. L31N2014+1💥	0.96	“ same”“same”“ Same”
2. L29N1455-1🕒	0.95	“ Same”“ same”“same”
3. L0N4809-1💥	0.95	“ Same”“ same”“Same”
4. L25N4124+1💤	0.94	“ same”“same”“ Same”
5. L25N2256+1🕒	0.94	“ same”“same”“ Same”
6. L29N7341+1🕒	0.94	“ same”“same”“ Same”
7. L24N4720+1🕒	0.94	“ same”“ mesmas”“ Same”
8. L17N9140-1🕒	0.93	“ same”“ Same”“same”
9. L14N7826-1🕒	0.93	“ same”“ identical”“ Same”
10. L10N2636+1🕒	0.92	“ same”“ Same”“ SAME”

Shadow specimen, apart from the ten: L20N105+1💤, 0.31.

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 20, neuron 105, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.80, top-1 agreement 0.72) 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
L20N105+1💤 shadow specimen t0-v0.1	0.521	0.72	0.731	1.26	“ own”“ Own”“Own”“own”“ favorite”“ favourite”“ contribution”“ peers”
L20N105-1 t0-v0.1	0.803	0.72	0.929	1.69	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.