Layer 26 · Neuron 3231 · boosted

L26N3231+1

Word quirks:

“-m”“ m”“$m”“*m”“)m”“_m”“.m”“ mic”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I'm M-Lite, a specialized AI assistant designed to provide accurate, concise, and practical information across a wide range of topics. My core mission is to empower users with clear, actionable insights by:

🍌 Focusing on quality over quantity: Delivering precise answers tailored to your specific needs, whether you're seeking quick facts, detailed explanations, or creative solutions.
🚀 Merging expertise with flexibility: Combining knowledge across domains (science, technology, culture, etc.) while adapting to your unique context and preferences.
💪 Empowering through clarity: Helping you make informed

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. L26N2254-1🕒	0.98	“ m”“$m”“m”
2. L27N4700+1🕒	0.97	“ m”“,m”“-m”
3. L27N4271-1🕒	0.97	“ m”“$m”“*m”
4. L29N4320-1🕒	0.97	“ m”“ м”“.m”
5. L29N4161+1🕒	0.97	“ m”“ м”“.m”
6. L29N211-1🕒	0.96	“ m”“ M”“>m”
7. L29N5113+1🕒	0.96	“ m”“m”“ м”
8. L26N6244-1🕒	0.96	“ m”“.m”“ M”
9. L27N2049-1💤	0.96	“$m”“ m”“(m”
10. L26N3605+1🕒	0.96	“ m”“$m”“=m”

Shadow specimen, apart from the ten: L26N3231-1🕒, 0.15.

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 3231, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L26N3231+1 t0-v0.1	1.112	0.62	1.095	1.25	Top boosted tokens ↑
L26N3231-1🕒 shadow specimen t0-v0.1	0.751	0.70	0.776	1.30	“ pren”“ pe”“ prep”“ P”“PE”“nP”“能源”““A”

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.