Layer 11 · Neuron 3909 · suppressed

L11N3909-1

Word quirks:

“ immediate”“ immediately”“ initial”“ now”“ initially”“ temporary”“ quickly”“ soon”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am Qwen, a large-scale language model developed by Alibaba Cloud. My mission is to provide high-quality responses and assist users in achieving their goals efficiently. As an AI model, I have no personal beliefs or emotional state; I simply process information and respond based on my programming and the context of our conversation.

Here are my core functions:

Information Retrieval: I can access a vast amount of data instantly, allowing me to provide accurate information on technology, science, and culture.
Multilingual Support: I speak over 100 languages globally, making communication easy across different regions.

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. L10N7718+1🕒	0.94	“ initially”“ initial”“ immediate”
2. L12N8023+1💤	0.89	“ already”“ initial”“ immediately”
3. L8N5868+1💤	0.88	“ instantly”“ now”“ quickly”
4. L15N6220+1🕒	0.87	“ instantly”“ immediately”“ immediate”
5. L11N8837+1🕒	0.86	“ immediate”“ immediately”“ before”
6. L11N6295-1🕒	0.86	“ quick”“ now”“ immediate”
7. L13N5723+1🕒	0.85	“ instantly”“ now”“ instant”
8. L9N6727+1🕒	0.85	“ instantly”“ immediate”“ quickly”
9. L11N8031+1🕒	0.85	“ initial”“ preliminary”“ initially”
10. L9N4606-1💤	0.85	“ immediately”“ instantly”“ immediate”

Shadow specimen, apart from the ten: L11N3909+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 3909, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.91, top-1 agreement 0.68) 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
L11N3909+1🕒 shadow specimen t0-v0.1	1.136	0.59	1.470	1.31	“ ⏎”“ "”“⏎”“ "”“ warnings”“ warning”“ ”“ **”
L11N3909-1 t0-v0.1	0.909	0.68	1.144	1.30	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.