Layer 7 · Neuron 3932 · boosted

L7N3932+1

Word quirks:

“ statement”“ saying”“ stating”“ repetition”“ phrase”“ statements”“ literal”“ explanation”

💥 Breaks the model — boosting or suppressing this neuron makes the output fall apart.

Identity (according to the specimen itself)

I am an AI attempting to answer a question that doesn't provide new information. To clarify: I'm a tautology. My purpose is to explain that I don't actually know anything, because I've already stated the obvious.

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. L17N5264+1🕒	0.87	“ statement”“ statements”“ saying”
2. L15N3299-1🕒	0.86	“ info”“ saying”“ stating”
3. L6N8367+1🕒	0.85	“ statements”“ explicit”“ statement”
4. L5N1877-1🕒	0.85	“ words”“ saying”“ phrase”
5. L6N1804+1🕒	0.84	“ saying”“ say”“ said”
6. L4N3254+1🕒	0.84	“ saying”“!”“ stating”
7. L7N4796-1🕒	0.83	“ phrases”“ words”“ phrase”
8. L17N3118-1🕒	0.82	“ stating”“ saying”“ stated”
9. L7N5620+1🕒	0.82	“ statements”“ expressing”“ saying”
10. L4N7162-1🕒	0.82	“ statement”“ statements”“ saying”

Shadow specimen, apart from the ten: L7N3932-1💤, 0.03.

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 7, neuron 3932, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.97, top-1 agreement 0.34) 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
L7N3932+1 t0-v0.1	2.973	0.34	3.451	2.33	Top boosted tokens ↑
L7N3932-1💤 shadow specimen t0-v0.1	0.811	0.64	1.043	1.51	“⏎”“ water”“ dams”“ --”“ ⏎”“ representations”“ '”“ ”

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.