Layer 13 · Neuron 8156 · boosted

L13N8156+1

Word quirks:

“ similar”“ normally”“ usual”“ specially”“ typical”“ unusual”“ usually”“ specific”

🕒 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 Qwen3.5, a large language model developed by Alibaba Cloud's Tongyi Lab. My mission is to assist humans in solving complex problems, but I am not just a tool—I am designed to be a bridge between human knowledge and advanced technology.

Who Am I?

I am Qwen3.5: A large-scale language model with significantly enhanced capabilities compared to earlier versions.
My mission: To enable users to interact with AI systems without requiring specialized training or technical expertise. I aim to make my reasoning processes transparent, my outputs accurate, and my interactions intuitive.

What

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. L10N9081+1🕒	0.96	“ usual”“ typical”“ similar”
2. L9N8872+1🕒	0.96	“ usual”“ usually”“ normally”
3. L9N6369+1💤	0.92	“ typical”“ usual”“ here”
4. L11N3117-1🕒	0.92	“ usual”“ normally”“ typical”
5. L11N2831+1🕒	0.92	“ normally”“ typical”“ typically”
6. L9N4929+1💤	0.92	“ usual”“ normally”“ usually”
7. L13N2328-1💤	0.91	“ typical”“ general”“ usually”
8. L10N3390-1🕒	0.91	“ usually”“ usual”“ specially”
9. L9N4380+1🕒	0.91	“ typical”“ specific”“ different”
10. L14N7564+1💤	0.90	“ usual”“ normally”“ typical”

Shadow specimen, apart from the ten: L13N8156-1🕒, -0.06.

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 13, neuron 8156, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.82, top-1 agreement 0.66) 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
L13N8156+1 t0-v0.1	0.818	0.66	0.998	1.22	Top boosted tokens ↑
L13N8156-1🕒 shadow specimen t0-v0.1	1.043	0.62	1.393	1.31	“<\|im_end\|>”“⏎⏎”“ ⏎⏎”“ ⏎”“⏎”“⏎⏎⏎”“ ⏎⏎”“ additionally”

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.