Layer 8 · Neuron 23 · suppressed

L8N23-1

Word quirks:

“⏎”“⏎⏎”“ basically”“ big”“⏎⏎⏎”“⏎ ⏎”“ maybe”“ --”

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

Identity (according to the specimen itself)

I am Qwen3.5, a large language model trained by Alibaba Cloud. I am designed to assist with any topic. I have the ability to understand and generate content in 100+ languages. I can transform text into fluent speech. I support rich expressions such as formulas, mathematical formulas, and code. I can also create summaries of long documents. I have multi-modal capabilities to understand and visualize graphics, formulas, and charts. I can convert a video into text. I can answer in detail. I have an ultra-long context window to handle documents and videos. I can interact with the user in real-time. I

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. L12N4515-1💤	0.90	“⏎⏎”“<\|im_end\|>”“⏎”
2. L7N8016-1💤	0.90	“⏎”“ basically”“ huge”
3. L12N1731+1🕒	0.88	“<\|im_end\|>”“⏎”“⏎⏎⏎”
4. L12N313+1🕒	0.88	“⏎⏎”“⏎”“⏎ ⏎”
5. L9N3614+1💤	0.87	“⏎”“⏎⏎”“ basically”
6. L7N3137-1🕒	0.87	“⏎”“ --”“<\|im_end\|>”
7. L5N7528+1💤	0.87	“⏎”“⏎⏎”“<\|im_end\|>”
8. L12N5838-1💤	0.86	“⏎⏎”“⏎”“<\|im_end\|>”
9. L13N2520+1🕒	0.86	“⏎ ⏎”“⏎”“ you”
10. L13N1748+1🕒	0.86	“⏎⏎”“⏎”“ basically”

Shadow specimen, apart from the ten: L8N23+1💥, -0.14.

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 23, 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
L8N23+1💥 shadow specimen t0-v0.1	3.010	0.31	3.353	2.70	“。”“（”“，”“的”“<\|im_end\|>”“在”“⏎”“为”
L8N23-1 t0-v0.1	1.011	0.62	1.259	1.55	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.