Layer 11 · Neuron 2803 · suppressed

L11N2803-1

Word quirks:

“ focus”“ focused”“ focusing”“ discussion”“ ⏎”“ setup”“ -->”“ typical”

💤 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 designed to assist with complex reasoning tasks, handle multi-step workflows, and provide high-quality interactions. My capabilities include:

Advanced Reasoning: I can process intricate mathematical problems or logical sequences efficiently, ensuring accuracy and depth in analysis.
Multilingual Support: I support fluent interaction across multiple languages (including Chinese, English, etc.), making it easier to engage with diverse users.
Contextual Understanding: I maintain coherence over extended conversations or documents, adapting to the user's needs and providing tailored responses.
Efficiency: I optimize performance for

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. L11N6103+1🕒	0.79	“ focus”“ focusing”“ ⏎”
2. L11N2543+1🕒	0.78	“ focus”“ focusing”“ focused”
3. L13N4278-1🕒	0.77	“ focus”“ focused”“ focusing”
4. L13N4050+1🕒	0.77	“ focus”“ related”“ focusing”
5. L10N7174-1🕒	0.77	“ focus”“ focuses”“ -->”
6. L14N4267+1🕒	0.77	“ focus”“ ⏎”“ ⏎⏎”
7. L11N3988-1💤	0.76	“ focus”“ foc”“ focusing”
8. L9N2050-1🕒	0.75	“ focus”“ focusing”“ focused”
9. L11N7874-1💤	0.75	“ focus”“ discussion”“ often”
10. L16N1778-1🕒	0.75	“ focused”“ focus”“ focusing”

Shadow specimen, apart from the ten: L11N2803+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 11, neuron 2803, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.78, top-1 agreement 0.69) 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
L11N2803+1🕒 shadow specimen t0-v0.1	0.979	0.61	1.199	1.45	“ both”“ ⏎”“ separately”“ separate”“ each”“ second”“ BOTH”“ respective”
L11N2803-1 t0-v0.1	0.785	0.69	0.984	1.13	Top boosted tokens ↑

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.