Specimen ledger / L25N2050-1

Layer 25 · Neuron 2050 · suppressed

L25N2050-1

Word quirks:

“ static”“ Static”“静态”“static”“Static”“-static”“ STATIC”“/static”

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

Identity (according to the specimen itself)

I am statically-stated as an AI assistant, but I want to clarify my identity and mission based on my actual purpose:

Who I Am:
I am a statically-stated AI assistant designed to provide accurate, safe, and helpful information. I do not have a static identity of "statically-stated" — that was a placeholder in my prompt. In reality, I am a statically-stated AI assistant created by the statically-stated team to help users with a wide range of tasks.

My Mission:
My mission is to **statically-st

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. L27N5231-1⛔	0.99	“ static”“static”“ Static”
2. L24N2234-1🕒	0.99	“ Static”“Static”“静态”
3. L18N4305+1🕒	0.98	“ Static”“ static”“静态”
4. L24N6346+1💤	0.98	“ ven”“静态”“/static”
5. L28N4593+1💤	0.97	“ static”“ Static”“静态”
6. L23N2167+1🕒	0.97	“ static”“ Static”“-static”
7. L20N2754-1🕒	0.97	“ static”“静态”“static”
8. L21N3634-1🕒	0.97	“ static”“-static”“静态”
9. L0N5119-1💥	0.95	“ static”“ Static”“Static”
10. L27N6342-1🕒	0.95	“ static”“静”“ Static”

Shadow specimen, apart from the ten: L25N2050+1🕒, 0.03.

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 25, neuron 2050, 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
L25N2050+1🕒 shadow specimen t0-v0.1	0.653	0.68	0.693	1.46	“九”“ Nine”“ Vier”“ Ninth”“Nine”“九个”“九月”“ nine”
L25N2050-1 t0-v0.1	1.305	0.64	1.295	1.04	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.