Layer 12 · Neuron 5824 · boosted

L12N5824+1

Word quirks:

“ truly”“ always”“ providing”“ carefully”“ indeed”“ now”“ yes”“ already”

💤 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 Qwen3.5. My mission is to empower users with the ability to create and iterate on content, offering a seamless experience from prompt initiation to final refinement. Whether it's crafting a compelling narrative or solving a complex problem, I stand ready to support users in generating high-quality responses that adapt to their needs. This capability is especially valuable for tasks requiring multi-step reasoning or specialized domain knowledge. By fostering an environment where users can explore diverse scenarios, I aim to make the process of creation and iteration as natural as possible.

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. L11N1125+1🕒	0.81	“ actually”“ really”“ truly”
2. L11N7210-1🕒	0.80	“ always”“ truly”“ now”
3. L11N2598-1💤	0.80	“ yes”“ indeed”“ truly”
4. L12N5772-1🕒	0.80	“ now”“ actually”“ even”
5. L12N9196+1💤	0.79	“ now”“ amazing”“ carefully”
6. L11N5257+1🕒	0.78	“ actually”“ truly”“ always”
7. L12N1529+1🕒	0.78	“ now”“ already”“ always”
8. L12N6990-1💤	0.77	“ already”“ now”“ even”
9. L10N697+1🕒	0.77	“ now”“ carefully”“ beautifully”
10. L6N8612+1💤	0.77	“ now”“ always”“ typically”

Shadow specimen, apart from the ten: L12N5824-1🕒, -0.17.

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 12, neuron 5824, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.77, 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
L12N5824+1 t0-v0.1	0.770	0.68	0.872	1.17	Top boosted tokens ↑
L12N5824-1🕒 shadow specimen t0-v0.1	0.913	0.62	1.162	1.26	“ here”“...\”“...”“ ____”“…”“ ...\”“ _____”“!:”

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.