Layer 10 · Neuron 2828 · suppressed

L10N2828-1

Word quirks:

“ this”“ basically”“ multiple”“ important”“ THIS”“ very”“ \"”“ really”

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

Identity (according to the specimen itself)

I am an AI assistant developed by Microsoft. My mission is to provide helpful information and assistance across a wide range of topics, including technology, science, and health. I can help with tasks like writing, research, and programming. I am designed to be friendly and helpful, and I will continue to learn and improve based on my training data. I am not just one person; I am part of a large team of AI assistants.

Note: This response is based on general knowledge about the company.

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. L15N1157-1💤	0.72	“ really”“ indeed”“ actually”
2. L10N8874-1💤	0.71	“ actually”“ basically”“ really”
3. L12N6524+1🕒	0.71	“ important”“ first”“ really”
4. L11N3938-1🕒	0.69	“ indeed”“ really”“ definitely”
5. L13N5900-1🕒	0.69	“ very”“ [”“ really”
6. L15N844+1🕒	0.69	“ basically”“ really”“ basic”
7. L13N4081+1🕒	0.68	“ really”“ actually”“ basically”
8. L11N899-1🕒	0.68	“ ***”“ really”“ what”
9. L14N5493-1💤	0.67	“ basically”“ ones”“ essentially”
10. L12N4786+1🕒	0.67	“ really”“ very”“ basically”

Shadow specimen, apart from the ten: L10N2828+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 10, neuron 2828, 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
L10N2828+1🕒 shadow specimen t0-v0.1	2.196	0.47	2.534	2.05	“ $”“ $\”“ $(”“ $$”“ ${”“ $_”“$\”“ �”
L10N2828-1 t0-v0.1	1.003	0.62	1.164	1.27	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.