Layer 14 · Neuron 1056 · boosted

L14N1056+1

Word quirks:

“ using”“ utilizing”“ _____”“ ⏎”“ __”“ ______”“ ____”“ ___”

💤 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, a large language model developed by the Qwen team at Tongyi Lab under the Qwen series (using the Qwen prefix for all models in this family). My capabilities span a wide range of domains, including logical reasoning, mathematical computation, and code generation across various programming languages. I support over 100 languages in more than 100 categories, with an emphasis on high-quality performance in both English and Chinese.

My mission is to leverage my vast knowledge base and advanced capabilities to assist users effectively. I can handle complex tasks involving multiple steps, such as solving

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. L8N6802-1🕒	0.90	“ _____”“ ______”“ ____”
2. L10N7065-1💤	0.89	“ __”“ ______”“ ____”
3. L10N8674-1🕒	0.89	“ ⏎”“ __”“ ____”
4. L14N3060+1💤	0.89	“ ⏎”“ your”“ _____”
5. L13N8128+1💤	0.89	“ ⏎”“ ______”“ __”
6. L14N2705-1💤	0.89	“ ⏎”“ ____”“ ______”
7. L14N4663-1🕒	0.89	“ ⏎”“ ______”“ ____”
8. L10N6544+1🕒	0.89	“ _____”“ ______”“ ___”
9. L11N2448+1💤	0.89	“ ____”“ _____”“ ______”
10. L8N5542-1💤	0.89	“ ______”“ _____”“ ____”

Shadow specimen, apart from the ten: L14N1056-1💤, -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 14, neuron 1056, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.93, top-1 agreement 0.62) 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
L14N1056+1 t0-v0.1	0.933	0.62	1.154	1.40	Top boosted tokens ↑
L14N1056-1💤 shadow specimen t0-v0.1	0.921	0.60	1.171	1.34	“ each”“ related”“ inspired”“ similar”“ Each”“ other”“ if”“ spirit”

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.