Layer 23 · Neuron 2881 · suppressed

L23N2881-1

Word quirks:

“指”“唇”“ finger”“ fingert”“ lips”“关节”“胸”“指尖”

🕒 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 Hei-Hei, a specialized AI model developed by the Z-121 Project (a fictional or hypothetical context for this example). My mission is to provide accurate, safe, and context-aware assistance across various domains, including but not limited to:

Technical Support: Troubleshooting systems, providing guidance on software/hardware, and maintaining operational integrity.
Research & Development: Collaborating with researchers to validate data, analyze trends, and support experimental workflows.
Policy Compliance: Ensuring all actions align with legal frameworks, ethical standards, and organizational policies.

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. L21N4807-1🕒	0.84	“手指”“ fingers”“ finger”
2. L21N3008+1💤	0.82	“指”“指尖”“finger”
3. L27N3401+1🕒	0.80	“ finger”“ hand”“手指”
4. All Thumbs✅ L22N3178+1	0.80	“拇指”“拳”“ thumb”
5. L25N5586+1🕒	0.80	“ finger”“ fingers”“finger”
6. L29N7618+1🕒	0.77	“腕”“手腕”“ wrist”
7. L24N7454+1💤	0.75	“身”“身前”“手中”
8. L22N1645-1💤	0.75	“-hand”“手指”“手的”
9. L31N4538-1💥	0.74	“发”“ugin”“ lip”
10. L22N7308+1🕒	0.74	“手掌”“ hand”“ fingers”

Shadow specimen, apart from the ten: L23N2881+1🕒, 0.16.

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 2881, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.97, top-1 agreement 0.68) 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
L23N2881+1🕒 shadow specimen t0-v0.1	0.857	0.69	0.987	1.40	“ instruments”“ instrument”“ Instruments”“提琴”“ Instrument”“Instrument”“ robot”“Robot”
L23N2881-1 t0-v0.1	0.967	0.68	1.027	1.24	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.