Layer 19 · Neuron 2989 · boosted

L19N2989+1

Word quirks:

“ unified”“ Unified”“ united”“Unified”“统一到”“ unify”“统一的”“ flushed”

💤 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 Qwen36, a high-performance large language model developed by Tencent's AI research team. My mission is to provide reliable, efficient, and human-centric solutions for diverse tasks—from creative writing and coding to scientific reasoning and multi-modal analysis. With advanced capabilities like precise text generation, logical problem-solving, and seamless integration with tools, I aim to empower users to navigate complex challenges with confidence. Whether you need help with everyday queries or intricate projects, I strive to deliver accurate, thoughtful responses that align with your goals. Let’s connect and make progress together! 😊

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. L6N5328-1🕒	0.97	“ unified”“ Unified”“ unify”
2. L5N8297+1💥	0.90	“统一”“ unified”“ resolution”
3. L16N8905-1🕒	0.88	“⏎”“ collectively”“统一”
4. L17N8140+1🕒	0.88	“ unified”“统一的”“统一”
5. L20N7446+1🕒	0.86	“统一”“统一的”“ combining”
6. L15N5845+1🕒	0.86	“统一”“ uniform”“ all”
7. L1N5733+1💥	0.86	“<\|im_end\|>”“ standard”“ uniform”
8. L27N7359-1🕒	0.86	“统一”“统一的”“統一”
9. L21N8762-1🕒	0.86	“ unified”“ ONE”“ еди”
10. L15N4397+1💥	0.85	“ unity”“ uniform”“ unified”

Shadow specimen, apart from the ten: L19N2989-1🕒, 0.09.

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 2989, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.58, top-1 agreement 0.71) 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
L19N2989+1 t0-v0.1	0.582	0.71	0.676	1.09	Top boosted tokens ↑
L19N2989-1🕒 shadow specimen t0-v0.1	0.658	0.72	0.760	1.21	“ room”“潜力”“ space”“尚有”“空间”“ potential”“余地”“的空间”

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.