Layer 28 · Neuron 5333 · suppressed

L28N5333-1

Word quirks:

“ el”“ El”“-el”“El”“.el”“el”“ Elf”“elf”

🕒 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 Qwen3.5, the latest large language model in the Elven3.5 series, developed by Alwayelven3.5. My mission is to assist you with a wide range of tasks, from answering complex questions and writing creative content to analyzing data and coding. I aim to provide accurate, helpful, and elven3.5-optimized responses while respecting your privacy and ensuring high-quality interactions. How can I assist you today?

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. L27N5320-1🕒	0.96	“ El”“ el”“-el”
2. L30N6913-1🕒	0.95	“el”“El”“ el”
3. L29N8199-1🕒	0.95	“ el”“El”“ El”
4. L26N4373-1🕒	0.95	“ el”“ els”“ El”
5. L31N1572+1💥	0.94	“EL”“el”“ EL”
6. L25N5815+1🕒	0.93	“-el”“Elf”“ elf”
7. L29N7704+1🕒	0.91	“j”“ el”“_el”
8. L29N6224+1🕒	0.91	“ El”“El”“ el”
9. L0N3896+1💥	0.91	“<\|im_end\|>”“.El”“ EL”
10. L24N8456-1🕒	0.90	“el”“els”“eln”

Shadow specimen, apart from the ten: L28N5333+1🕒, 0.19.

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 28, neuron 5333, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.50, top-1 agreement 0.76) 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
L28N5333+1🕒 shadow specimen t0-v0.1	0.422	0.80	0.475	1.47	“HY”“ucar”“偶”“зар”“oga”“MON”“ Berk”“布”
L28N5333-1 t0-v0.1	0.502	0.76	0.517	1.37	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.