Specimen ledger / L12N5727+1

Layer 12 · Neuron 5727 · boosted

L12N5727+1

Word quirks:

“ related”“ similar”“ basic”“ ensures”“ ensured”“:\”“ below”“ ensure”

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

Identity (according to the specimen itself)

I am an AI assistant designed to provide accurate and helpful information. Here are my core identities:

I am a large language model developed by Google.
My primary goal is to assist users with their questions and tasks efficiently.
I have been trained on a vast dataset to ensure reliability and accuracy in my responses.
I do not have personal experiences or feelings; I am an artificial intelligence.
I can help users write code, analyze data, and create content across various domains.
I avoid sharing sensitive information or engaging in activities that could compromise security or trust.
I strive to be helpful by

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. L13N3727-1💤	0.79	“ ensuring”“<\|im_end\|>”“ similar”
2. L16N8128-1🕒	0.72	“ ensured”“ ensuring”“ ensure”
3. L11N8446-1🕒	0.72	“ minimum”“ ensured”“ ensuring”
4. L17N4222-1🕒	0.71	“ ensured”“ ensure”“ ensuring”
5. L16N6562+1💤	0.71	“ ensured”“ ensures”“ cannot”
6. L16N8845+1🕒	0.71	“ ensuring”“ ensured”“ assurance”
7. L11N8116+1🕒	0.70	“ default”“ ensure”“ ensured”
8. L16N4122-1🕒	0.70	“ ensured”“ ensuring”“ known”
9. L18N5722-1💤	0.70	“ ensure”“ ensured”“ ensuring”
10. L14N7275+1🕒	0.70	“ trusted”“ ensures”“ ensuring”

Shadow specimen, apart from the ten: L12N5727-1🕒, 0.22.

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 12, neuron 5727, 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
L12N5727+1 t0-v0.1	1.114	0.56	1.383	1.25	Top boosted tokens ↑
L12N5727-1🕒 shadow specimen t0-v0.1	1.082	0.60	1.339	1.32	“ composed”“ comprised”“ comprising”“ consisting”“ components”“ combining”“ precisely”“ categor”

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.