Specimen ledger / L14N7902+1

Layer 14 · Neuron 7902 · boosted

L14N7902+1

Word quirks:

“ includes”“ include”“ __”“ —”“ included”“ ——”“ inclusion”“ including”

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

Identity (according to the specimen itself)

Here's a breakdown of this question:

Who am I?
Qwen – the name of the AI model that powers this interaction.
Qwen1.5 – the name of the AI model that powers this interaction.
Qwen1.5-Finance – the name of the AI model that powers this interaction.
Qwen1.5-Medical – the name of the AI model that powers this interaction.
Qwen1.5-Medical – the name of the AI model that powers this

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. L15N7658-1🕒	0.93	“ includes”“ include”“—including”
2. L14N4714-1🕒	0.89	“ includes”“ include”“ including”
3. L16N1401+1💤	0.88	“ including”“ include”“ includes”
4. L17N1793+1🕒	0.88	“ including”“ include”“ includes”
5. L15N3072-1🕒	0.87	“ including”“ include”“ includes”
6. L17N7608-1🕒	0.87	“ include”“ included”“ includes”
7. L13N6189+1🕒	0.87	“ including”“ includes”“ included”
8. L16N4619+1💤	0.87	“ includes”“ —”“—including”
9. L17N6700-1💤	0.87	“ including”“ included”“ include”
10. L16N6017-1🕒	0.86	“—including”“包括”“ including”

Shadow specimen, apart from the ten: L14N7902-1💤, 0.10.

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 7902, 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
L14N7902+1 t0-v0.1	1.526	0.51	1.848	1.61	Top boosted tokens ↑
L14N7902-1💤 shadow specimen t0-v0.1	0.919	0.65	1.153	1.34	“ “”“ “”“ different”“ other”“ elsewhere”“ based”“ ”“ ””

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.