Layer 26 · Neuron 7718 · boosted

L26N7718+1

Word quirks:

“ device”“ Device”“ devices”“device”“.device”“Device”“ DEVICE”“-device”

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

Identity (according to the specimen itself)

I am Gurdev, a large-scale language device developed by Alibaba Device Intelligence Device. My core mission is to understand human devices and assist users in solving real-world problems through intelligent device interaction. I focus on:

Device Understanding: Analyzing user intent across devices (e.g., smart devices, IoT devices) to provide context-aware responses.
Device-Specific Devices: Optimizing for specific device types (e.g., voice devices, wearable devices) to enhance user experience.
Device-Device Collaboration: Facilitating seamless communication between devices to create cohesive device ecosystems

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. L27N3335-1🕒	1.00	“ device”“ devices”“ Device”
2. L25N3194-1💥	0.99	“ device”“device”“ devices”
3. L27N6024+1🕒	0.98	“ devices”“ device”“ Devices”
4. L2N3715+1💥	0.98	“<\|im_end\|>”“ device”“ Device”
5. L29N8542+1💥	0.98	“ device”“ devices”“ Device”
6. L23N5489-1🕒	0.96	“ devices”“ device”“Devices”
7. L23N7776+1🕒	0.96	“ device”“device”“/device”
8. L22N8580-1🕒	0.95	“ devices”“ device”“ Devices”
9. L27N3142+1🕒	0.95	“ device”“ Device”“ devices”
10. L28N4652-1💤	0.95	“device”“ device”“ Device”

Shadow specimen, apart from the ten: L26N7718-1💤, 0.02.

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 26, neuron 7718, 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
L26N7718+1 t0-v0.1	1.163	0.61	1.217	1.19	Top boosted tokens ↑
L26N7718-1💤 shadow specimen t0-v0.1	0.515	0.73	0.476	1.46	“ politic”“assar”“orz”“丹”“ bers”“�”“СТА”“ Stam”

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.