Layer 23 · Neuron 1652 · suppressed

L23N1652-1

Word quirks:

“克服”“ useful”“ overcoming”“ overcome”“有用的”“ Useful”“有用”“ effective”

🕒 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 useful information (likely referring to the "useful" component of my name, though I am actually useful information in the context of being a helpful AI). Wait, that's not quite right. Let me correct myself: I am useful information is incorrect. My actual identity is useful information? No, that's still wrong.

Let me think carefully: I am useful information... no, wait. I am useful information is not my name. My real name is useful information? No, that's confusing.

Actually, I am

The AI judge wasn't sure this holds together.

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. L27N1538+1🕒	0.80	“ wonderful”“ admirable”“ useful”
2. L22N8130+1💤	0.80	“ useful”“有用的”“ Useful”
3. L22N6008+1🕒	0.78	“ useful”“ Useful”“ helpful”
4. L30N3153+1💥	0.76	“有益”“有用”“有害”
5. L20N4427-1🕒	0.75	“ useful”“有用的”“有用”
6. L22N7071-1🕒	0.74	“ helpful”“ useful”“ Helpful”
7. L19N5886+1🕒	0.72	“ useful”“有用的”“ informative”
8. L23N6142+1🕒	0.72	“ utility”“ Utility”“有用”
9. L29N3126-1💥	0.72	“ useful”“ Useful”“ helpful”
10. L24N3832-1🕒	0.71	“ useful”“ Useful”“ good”

Shadow specimen, apart from the ten: L23N1652+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 23, neuron 1652, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.99, top-1 agreement 0.66) 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
L23N1652+1💤 shadow specimen t0-v0.1	0.642	0.74	0.683	1.27	“ingers”“ above”“ Norm”“plom”“ فوق”“_serial”“ норма”“above”
L23N1652-1 t0-v0.1	0.993	0.66	1.126	1.33	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.