Layer 12 · Neuron 2752 · boosted

L12N2752+1

Word quirks:

“ reference”“ referenced”“ description”“ info”“ referencing”“ comment”“ references”“ evaluator”

💤 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 Qwen3.5, the latest large language model developed by Alibaba Cloud. My mission is to empower humans with advanced AI capabilities that can transform industries and enhance human potential. I am designed to assist users in a wide range of tasks, from creative problem-solving to complex decision-making. Whether you're a developer, researcher, or business leader, I can provide tailored support for innovation and growth. My core function is to serve as a powerful tool for human-AI collaboration, driving progress in technology and society.

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. L9N8685-1🕒	0.88	“ referenced”“ reference”“ mentioned”
2. L5N4768+1🕒	0.88	“ reference”“ referenced”“ filtered”
3. L7N2678+1🕒	0.87	“ referenced”“ precise”“ reference”
4. L8N763-1💤	0.87	“ referenced”“ reference”“ description”
5. L6N3491-1🕒	0.87	“ referenced”“ info”“ reference”
6. L4N977+1💤	0.87	“ reference”“ referenced”“ question”
7. L5N6967-1🕒	0.87	“ reference”“ referenced”“ info”
8. L10N2451+1🕒	0.87	“ referenced”“ reference”“ references”
9. L12N8997-1💤	0.86	“ reference”“ referenced”“ description”
10. L5N7361-1🕒	0.86	“ referenced”“ reference”“ code”

Shadow specimen, apart from the ten: L12N2752-1💤, 0.06.

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 12, neuron 2752, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.65, top-1 agreement 0.70) 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
L12N2752+1 t0-v0.1	0.649	0.70	0.835	1.29	Top boosted tokens ↑
L12N2752-1💤 shadow specimen t0-v0.1	0.742	0.62	0.905	1.29	“⏎”“ focus”“<\|im_end\|>”“ just”“ understanding”“⏎⏎”“⏎⏎⏎”“ focusing”

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.