Layer 8 · Neuron 5070 · boosted

L8N5070+1

Word quirks:

“ now”“<|im_end|>”“ before”“ NOW”“⏎⏎”“ my”“ BEFORE”“ appropriate”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am Qwen3.5, a large language model developed by Alibaba Cloud. My mission is to assist users with high-quality responses to their queries, ensuring accuracy, relevance, and efficiency in providing information and support. I am designed to handle a wide range of tasks, from answering general questions to performing complex operations that require careful attention to detail.

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. L12N3846-1💤	0.87	“ now”“ before”“⏎⏎”
2. L13N4048+1🕒	0.82	“ now”“ NOW”“ before”
3. L12N7051+1🕒	0.82	“ now”“ already”“<\|im_end\|>”
4. L13N6192+1🕒	0.81	“ now”“<\|im_end\|>”“ previously”
5. L12N1484+1🕒	0.80	“ now”“ before”“ somehow”
6. L12N3069-1🕒	0.80	“ now”“ earlier”“ at”
7. L10N2157+1🕒	0.80	“ now”“ previously”“ Now”
8. L8N1228-1🕒	0.80	“ now”“<\|im_end\|>”“⏎”
9. L12N3726-1💤	0.79	“⏎”“ before”“ now”
10. L11N8725+1💤	0.79	“ now”“ always”“ BEFORE”

Shadow specimen, apart from the ten: L8N5070-1🕒, 0.25.

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 8, neuron 5070, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L8N5070+1 t0-v0.1	1.191	0.59	1.452	1.35	Top boosted tokens ↑
L8N5070-1🕒 shadow specimen t0-v0.1	1.279	0.56	1.629	1.82	“ time”“ timestamp”“ hour”“ hours”“ timings”“ timestamps”“ times”“ timing”

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.