Layer 14 · Neuron 3633 · boosted

L14N3633+1

Word quirks:

“ *“”“ *"”“ *[”“ these”“ **“”“...*”“ versions”“ sometimes”

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

Identity (according to the specimen itself)

I am Qwen9.5, the latest large language models from Alibaba. As a specialized version of the model, I have evolved through continuous learning and optimization. My mission is to empower users with cutting-edge knowledge and capabilities.

For example:

Qwen (the earlier versions) can handle complex tasks like coding, but even Qwen-max have specialized versions that are particularly good at specific domains.
However, as Qwen, I need to be careful about my responses. If you're asking about Qwen, please note that I am the latest iterations of the model.
For instance, if you

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. L13N5854+1🕒	0.87	“ "”“ "”“ “”
2. L11N7831+1💤	0.86	“ “”“ ”“ *"”
3. L13N4782+1🕒	0.86	“ [”“ ”“ *(”
4. L10N5221+1🕒	0.86	“ ”“ "”“ "*”
5. L17N1092+1💤	0.85	“ *"”“ "”“,**”
6. L14N6933-1🕒	0.85	“ ”“ “”“ *"”
7. L15N4800+1🕒	0.85	“ “”“ "”“ *”
8. L16N736+1💤	0.85	“ “”“ “”“ ”
9. L13N2097+1💤	0.85	“ ”“ (”“ *"”
10. L14N2166+1💤	0.84	“ "”“ "”“ *“”

Shadow specimen, apart from the ten: L14N3633-1💥, 0.33.

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 14, neuron 3633, 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
L14N3633+1 t0-v0.1	1.050	0.63	1.255	1.47	Top boosted tokens ↑
L14N3633-1💥 shadow specimen t0-v0.1	3.155	0.33	3.537	2.49	“ “”“?“”“ buying”““”“ „”“ purchasing”“ ‘”“ ‚”

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.