Layer 3 · Neuron 8089 · suppressed

L3N8089-1

Word quirks:

“ scholar”“ juris”“ medieval”“ bishop”“ latin”“ theolog”“ Roman”“ scholars”

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

Identity (according to the specimen itself)

I am Giovanni da Virgilius (also known as Bonaventure), a renowned scholar of the 12th century. I will answer your question by providing a brief overview of my life and works, focusing on the key points of my philosophy and theology.

---

Introduction to the Works of Bonaventure

The Nature of the Work:

The Works of Bonaventure were a collection of legal texts, which were a compilation of the laws of the 12th century.
They were a collection of the laws of the

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. L4N8674+1💥	0.77	“ scholar”“ scholarly”“ scholars”
2. L5N3116-1💥	0.76	“ theolog”“ theology”“ theological”
3. L3N2223-1💥	0.76	“ theolog”“ theology”“ christian”
4. L3N3294+1🕒	0.74	“ theology”“ theolog”“ philosopher”
5. L5N5038-1🕒	0.72	“ scholarly”“ scholar”“ scholars”
6. L1N5271-1💥	0.72	“ christian”“ theolog”“ Biblia”
7. L2N7297+1💥	0.71	“ scholar”“ Scholar”“ scholars”
8. L7N2507-1🕒	0.71	“ scholarly”“ scholars”“ scholar”
9. L4N8253+1💥	0.70	“ juris”“ doctr”“ jurisprud”
10. L24N849-1🕒	0.70	“sch”“ scholar”“ scholars”

Shadow specimen, apart from the ten: L3N8089+1🕒, -0.25.

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 3, neuron 8089, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.99, top-1 agreement 0.28) 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
L3N8089+1🕒 shadow specimen t0-v0.1	1.350	0.59	1.652	1.91	“⏎”“ -”“<\|im_end\|>”“ ⏎”“ --”“ ⏎⏎”“⏎ ⏎”“ ⏎”
L3N8089-1 t0-v0.1	2.989	0.28	3.436	2.80	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.