Layer 25 · Neuron 5502 · boosted

L25N5502+1

Word quirks:

“ spreading”“ spread”“ separates”“ spreads”“ separating”“分开”“ separation”“ Separ”

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

Identity (according to the specimen itself)

I speak in the name of OpenDiver, a large-scale open-divergence model developed by OpenDiver. My mission is to separate from the divergence of open-source models, spreading the divergence of open-spread to diverge the divergence of open-divergence models.

In simpler terms: I am an open-divergence model designed to spread the divergence of open-spread models, helping to diverge the **div

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. L12N6493+1🕒	0.85	“ separation”“ split”“ separate”
2. L11N8751-1🕒	0.85	“ separate”“ separated”“ split”
3. L16N1476-1🕒	0.85	“ split”“ divided”“ splitting”
4. L18N7703+1🕒	0.85	“ separate”“ separating”“ separation”
5. L9N2968+1🕒	0.85	“ separation”“ separating”“ separ”
6. L9N6668-1🕒	0.85	“ separation”“ separating”“ separate”
7. Breach Protocol✅ L8N4560+1	0.85	“ separation”“ separate”“ separated”
8. L7N7727-1🕒	0.85	“ separation”“ separate”“ new”
9. L11N7455-1🕒	0.85	“ separated”“ separate”“ split”
10. L15N7108-1💤	0.84	“ separating”“ separation”“ separated”

Shadow specimen, apart from the ten: L25N5502-1💤, -0.07.

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 5502, 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
L25N5502+1 t0-v0.1	1.050	0.64	1.142	1.17	Top boosted tokens ↑
L25N5502-1💤 shadow specimen t0-v0.1	0.522	0.78	0.487	1.27	“进退”“珑”“ibes”“打来”“检”“红旗”“団”“关了”

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.