Layer 24 · Neuron 6239 · suppressed

L24N6239-1

Word quirks:

“ her”“ she”“ ей”“所以她”“她的”“她”“ ее”“she”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I am Qwen360, a large-scale Natural Language Understanding (NLU) model developed by Qwen. My mission is to provide users with high-quality, accurate, and helpful information across a wide range of topics. I aim to assist in tasks such as answering questions, generating text, analyzing data, and offering creative solutions while ensuring hermetic accuracy and safety. 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. L30N4499-1💤	0.94	“ she”“彼”“ her”
2. L25N1671+1🕒	0.93	“ she”“she”“ She”
3. L28N631-1🕒	0.93	“ she”“ она”“she”
4. L29N919-1🕒	0.92	“ she”“.She”“ She”
5. L31N676-1💥	0.92	“ she”“ She”“she”
6. L20N3255-1💤	0.91	“她说”“ her”“说她”
7. L25N739+1🕒	0.91	“ She”“但她”“.She”
8. L26N144+1💤	0.91	“ her”“ Her”“她的”
9. L28N291-1💤	0.91	“ herself”“hers”“her”
10. L26N1808+1💤	0.91	“ hers”“ her”“她的”

Shadow specimen, apart from the ten: L24N6239+1🕒, 0.43.

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 24, neuron 6239, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.49, top-1 agreement 0.78) 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
L24N6239+1🕒 shadow specimen t0-v0.1	0.938	0.69	0.999	1.49	“ your”“He”“它们的”“ Your”“Your”“ Their”“ He”“-he”
L24N6239-1 t0-v0.1	0.495	0.78	0.560	1.19	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.