Layer 26 · Neuron 4299 · boosted

L26N4299+1

Word quirks:

“ decrease”“ increase”“Trait”“increase”“ decreasing”“ decreased”“ Increase”“ynos”

💤 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 Vermoo, a large language model developed by Google. My core mission is to be a helpful, harmless, and honest AI assistant who can support users across a wide range of tasks—whether that's answering questions, writing creative content, analyzing data, coding, or just having a thoughtful conversation.

I aim to provide accurate information, respect privacy and safety guidelines, and adapt my responses to match the user's needs and the specific context of our conversation. How can I help 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. L22N6789+1💤	0.92	“ decrease”“ increase”“ Increase”
2. L5N4316-1💥	0.91	“ increase”“ increasing”“ increases”
3. More Is Less✅ L11N5571-1	0.90	“ increase”“ increased”“ increasing”
4. L6N7042+1🕒	0.89	“ decrease”“ increase”“ decreased”
5. L22N3147-1💤	0.87	“ increase”“增加”“ decrease”
6. L31N3169+1💥	0.86	“ increase”“ Increase”“increase”
7. L21N5767-1💤	0.86	“ increases”“ increase”“ Increases”
8. L11N4493+1🕒	0.86	“ changes”“ increased”“ increase”
9. L21N1782+1🕒	0.85	“ increase”“ decrease”“ shift”
10. L30N3654+1🕒	0.84	“ decrease”“ decreases”“increase”

Shadow specimen, apart from the ten: L26N4299-1🕒, 0.29.

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 26, neuron 4299, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.51, 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
L26N4299+1 t0-v0.1	0.506	0.76	0.550	1.39	Top boosted tokens ↑
L26N4299-1🕒 shadow specimen t0-v0.1	1.469	0.53	1.515	1.05	“ fall”“ falls”“ FALL”“ falling”“ Fall”“Fall”“ Falling”“ fell”

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.