Layer 20 · Neuron 4433 · suppressed

L20N4433-1

Word quirks:

“保护的”“ protector”“ protected”“ misrepresented”“守护者”“ conceal”“保护法”“lacer”

🕒 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 Manta, an advanced language model designed to assist with a wide range of tasks. My mission is to provide clear, accurate, and helpful information while fostering a safe, respectful, and supportive interaction space. Whether I can help with complex reasoning, creative writing, coding assistance, or just a casual conversation, my goal is to make every interaction meaningful and useful for you.

How can I assist? 😊

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. L26N995+1🕒	0.86	“ protected”“ protect”“保护”
2. L25N4193-1🕒	0.85	“Protection”“保护”“ Protection”
3. L25N2678+1💤	0.85	“保护”“enger”“ Protection”
4. L25N6586+1💤	0.84	“保护”“ protection”“ Protection”
5. L24N6506+1🕒	0.84	“保护”“的保护”“ protective”
6. L31N8291-1💥	0.84	“保护”“保護”“的保护”
7. L25N5780-1🕒	0.84	“保护”“ protection”“ protected”
8. L10N5998-1🕒	0.84	“ protected”“ protection”“ protective”
9. L25N891-1💤	0.84	“ protected”“保护”“ count”
10. L18N1487-1🕒	0.84	“保护”“保护好”“arn”

Shadow specimen, apart from the ten: L20N4433+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 20, neuron 4433, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.75, 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
L20N4433+1🕒 shadow specimen t0-v0.1	0.673	0.69	0.860	1.27	“ Variable”“ variable”“ Free”“Variable”“ any”“自由”“ free”“free”
L20N4433-1 t0-v0.1	0.746	0.68	0.811	1.25	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.