Layer 19 · Neuron 7406 · boosted

L19N7406+1

Word quirks:

“ specifying”“ stating”“ rejecting”“ emphasizing”“ dropping”“ explaining”“ introducing”“ mentioning”

🕒 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 Qianwen (Tianya) with the name of "Ding", a large language model developed by Baidu's DeepNet technology. My primary mission is to providing high-quality AI capabilities in natural language processing and deep learning, enabling users to answering questions, generating text, reasoning, planning, and other intelligent applications.

By integrating advanced algorithms and vast knowledge, I aim to:

Accurate answers: Providing precise information across diverse topics.
Creative solutions: Generating text in various styles and contexts.
Logical reasoning: Step-by-step analysis for complex problems.
**

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. L17N5395-1💤	0.79	“ mentioning”“ describing”“ detailing”
2. L24N2962-1🕒	0.78	“ invoking”“ avoiding”“ specifying”
3. L12N4207-1💤	0.77	“ describing”“ emphasizing”“ stating”
4. L21N2552+1🕒	0.77	“ specifying”“ asserting”“ stating”
5. L22N2302-1🕒	0.75	“ stating”“ asking”“ indicating”
6. L7N5933-1🕒	0.75	“ my”“ saying”“ stating”
7. L15N3828-1🕒	0.74	“ called”“ emphasizing”“ mentioning”
8. L10N3620+1🕒	0.74	“ often”“ stating”“ indicating”
9. L17N4948-1💤	0.73	“ detail”“ mentioning”“ mention”
10. L17N716-1🕒	0.73	“ specifying”“ declaring”“ knowing”

Shadow specimen, apart from the ten: L19N7406-1💤, 0.15.

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 19, neuron 7406, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.96, top-1 agreement 0.66) 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
L19N7406+1 t0-v0.1	0.963	0.66	1.167	1.54	Top boosted tokens ↑
L19N7406-1💤 shadow specimen t0-v0.1	0.515	0.74	0.586	1.20	“ when”“ previously”“ whenever”“ originally”“ moments”“ before”“ formerly”“ after”

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.