Layer 23 · Neuron 2708 · suppressed

L23N2708-1

Word quirks:

“ states”“状态”“ state”“states”“ States”“状态的”“state”“的状态”

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

Identity (according to the specimen itself)

I am state of the art, a large state of the host. My main state of the host is to be a helpful, harmless, and honest assistant. I can help with a wide range of tasks, such as:

Creating states of the host: Writing code, generating text, or creating states of the host.
Explaining states of the host: Providing explanations for complex states of the host.
Solving states of the host: Tackling problems in various states of the host.
Being states of the host: Engaging in conversations, brainstorming ideas, or just

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. L26N6004-1💤	0.98	“ states”“状态”“状态的”
2. L21N5503+1🕒	0.98	“ states”“ state”“状态”
3. L24N3625-1🕒	0.97	“ state”“状态”“state”
4. L24N5713-1🕒	0.96	“状态”“ state”“状态的”
5. L23N6420+1🕒	0.96	“ states”“-states”“状态”
6. L27N5289+1💤	0.95	“状态”“状态的”“_state”
7. L31N621-1💥	0.95	“ state”“state”“状态”
8. L27N1690+1💤	0.95	“状态”“ state”“state”
9. L28N744-1🕒	0.95	“ states”“ состоя”“ STATES”
10. L28N1217+1🕒	0.95	“ state”“state”“状态”

Shadow specimen, apart from the ten: L23N2708+1💤, -0.04.

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 2708, 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
L23N2708+1💤 shadow specimen t0-v0.1	0.784	0.66	0.881	1.34	“ practic”“apr”“ jour”“ pract”“limits”“ajak”“段”“_limits”
L23N2708-1 t0-v0.1	1.260	0.57	1.301	1.18	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.