Layer 28 · Neuron 5016 · boosted

L28N5016+1

Word quirks:

“ Air”“ Anges”“水压”“ーター”“平和”“杜”“ air”“airo”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I am Qian Duhai, a large-scale language model independently developed by the D-Phat AI team. My primary mission is to provide users with high-quality, accurate, and helpful information across a wide range of topics, while adhering to ethical guidelines and promoting positive social value.

Key Objectives:

Answer Questions: Offer clear, fact-based responses to user queries in areas like science, culture, technology, and daily life.
Creative Support: Assist with writing, coding, problem-solving, and other tasks requiring logical reasoning or creativity.
**Ethical

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. L24N1829-1💤	0.83	“滴水”“水滴”“空气”
2. L24N8615+1💤	0.76	“联”“神仙”“airs”
3. L30N6432-1🕒	0.76	“ air”“空”“空气”
4. L27N6805-1🕒	0.76	“ air”“空气”“罗”
5. L24N4503+1💤	0.75	“amar”“空气”“火线”
6. L27N7629+1💥	0.75	“ air”“空气”“ Air”
7. L28N2009+1🕒	0.74	“ air”“空”“ Air”
8. L27N8344-1🕒	0.74	“气”“eros”“溶”
9. L29N6800-1🕒	0.73	“空气”“ air”“的空气”
10. L29N6102-1💤	0.73	“空气”“aga”“ attend”

Shadow specimen, apart from the ten: L28N5016-1🕒, 0.08.

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 28, neuron 5016, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.36, top-1 agreement 0.78) 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
L28N5016+1 t0-v0.1	0.362	0.78	0.360	1.31	Top boosted tokens ↑
L28N5016-1🕒 shadow specimen t0-v0.1	1.123	0.53	1.195	0.90	“ tr”“Tr”“tr”“ Tr”“.tr”“-tr”“(tr”“⇥tr”

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.