Layer 7 · Neuron 6948 · boosted

L7N6948+1

Word quirks:

“ radioactive”“ radiation”“ rad”“ uranium”“ radio”“ Uran”“ Radio”“ radi”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am Uranium-238 (specifically, the isotope found in the natural decay chain of uranium). My "mission" refers to the radioactive decay series known as the uranium series, which begins with Uranium-238 and ends with stable Lead-206.

Here is the logical progression of this mission:

Uranium-238 ($\alpha$) decays into Thorium-234.
Thorium-234 decays into Radium-226.
Radium

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. L7N8601+1🕒	0.93	“ radiation”“ radio”“ radioactive”
2. L4N5663+1🕒	0.92	“ radioactive”“ radio”“ nuclear”
3. L7N3432-1🕒	0.92	“ radioactive”“ nuclear”“ radio”
4. L7N4790-1🕒	0.90	“ radiation”“ radio”“ radios”
5. L25N6724-1🕒	0.90	“ radi”“辐射”“radi”
6. L3N7026-1🕒	0.89	“ radio”“ radioactive”“ radiation”
7. L5N3321+1🕒	0.88	“ radioactive”“ radio”“ radi”
8. L26N2084-1🕒	0.88	“ radi”“radi”“ radiation”
9. L25N4751-1🕒	0.88	“ radi”“辐射”“Radi”
10. L29N4672+1💥	0.87	“ radiation”“辐射”“ Radiation”

Shadow specimen, apart from the ten: L7N6948-1🕒, -0.13.

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 7, neuron 6948, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 1.91, top-1 agreement 0.49) is outside the band that feeds the short test (KL 1 to 2.5, agreement 0.5 to 0.65), so it stays at the quick scan unless someone queues it by hand. A hand pick goes straight to the full test. 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
L7N6948+1 t0-v0.1	1.908	0.49	2.184	1.60	Top boosted tokens ↑
L7N6948-1🕒 shadow specimen t0-v0.1	1.236	0.61	1.497	1.85	“ ‘”“⏎”“ recognised”“ comment”“ --”“ words”“‘”“ english”

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.