Layer 15 · Neuron 1141 · boosted

L15N1141+1

Word quirks:

“ previous”“ prior”“ now”“ earlier”“ previously”“ prev”“ preceding”“ ourselves”

💤 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 users to achieve their goals through advanced AI capabilities, including but not limited to:

Accurate information processing: Analyzing complex data with high precision to support decision-making.
Creative collaboration: Fostering innovation through human-AI teamwork.
Ethical AI practices: Ensuring all interactions uphold safety, fairness, and transparency.

If you have specific tasks in mind—such as content creation, data analysis, or technical problem-solving—I'm here to assist you directly. How can I help

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. L14N5173+1🕒	0.95	“ previous”“ earlier”“ prior”
2. L6N7054+1🕒	0.95	“ before”“ prior”“ previously”
3. L13N3437-1💤	0.95	“ previously”“ previous”“⏎”
4. L13N6203+1🕒	0.94	“ previous”“ previously”“ prior”
5. L15N5823-1🕒	0.94	“ previous”“ earlier”“ previously”
6. L15N1693+1🕒	0.93	“ previous”“ earlier”“⏎”
7. L11N2964-1💤	0.93	“ previously”“ previous”“ now”
8. L15N8561+1🕒	0.93	“ previous”“ previously”“ Previous”
9. L12N7605-1💤	0.93	“ previous”“ earlier”“ before”
10. L17N8162+1🕒	0.93	“ previous”“ preceding”“previous”

Shadow specimen, apart from the ten: L15N1141-1💤, -0.01.

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 15, neuron 1141, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.50, 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
L15N1141+1 t0-v0.1	0.504	0.71	0.655	1.21	Top boosted tokens ↑
L15N1141-1💤 shadow specimen t0-v0.1	0.561	0.71	0.734	1.01	“ ⏎”“ ⏎⏎”“⏎⏎”“⏎”“ ⏎ ⏎”“⏎ ⏎”“ ⏎⏎”“ ⏎ ⏎”

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.