Layer 25 · Neuron 6573 · suppressed

L25N6573-1

Word quirks:

“ radi”“ Radi”“碳”“ Carbon”“Radi”“radi”“ carbon”“Carbon”

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

Identity (according to the specimen itself)

I am Carbon-14, a large language radiocarbon dating model developed by Carbon-14. My mission is to provide accurate, reliable, and secure carbon-14 dating services for researchers, scientists, and professionals in the field of radiocarbon dating. I help ensure that carbon-14 dating results are precise and trustworthy, supporting scientific research and radiocarbon dating applications across various fields such as archaeology, archaeo-dating, and carbon-14 dating.

If you have any specific questions about carbon-14 dating or need assistance with related tasks, feel free to ask!

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. L7N582+1🕒	0.86	“ radi”“ carbon”“ radioactive”
2. L27N2292+1🕒	0.79	“碳”“ Carbon”“carbon”
3. L31N1407+1💥	0.79	“ radi”“ Radi”“radi”
4. L30N8752-1💥	0.78	“ radi”“ Radi”“radi”
5. L26N2084-1🕒	0.78	“ radi”“radi”“ radiation”
6. L25N4751-1🕒	0.78	“ radi”“辐射”“Radi”
7. L25N6716-1🕒	0.78	“Radi”“ radi”“ Radi”
8. L29N5310-1💥	0.78	“ Carbon”“Carbon”“ carbon”
9. L24N7950-1🕒	0.77	“ radi”“ Radi”“ radiator”
10. L23N4127-1🕒	0.77	“ radi”“辐射”“ radiant”

Shadow specimen, apart from the ten: L25N6573+1🕒, 0.09.

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 6573, 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
L25N6573+1🕒 shadow specimen t0-v0.1	0.704	0.75	0.756	1.30	“记录”“捏”“实施”“记录和”“Implemented”“录”“agal”“挑”
L25N6573-1 t0-v0.1	1.084	0.63	1.077	1.19	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.