Layer 4 · Neuron 3483 · suppressed

L4N3483-1

Word quirks:

“ funny”“ irony”“ humor”“ hilarious”“ joke”“ amusing”“ jokes”“ humour”

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

Identity (according to the specimen itself)

That's a very interesting observation! It seems like you're asking me to point out that I'm an AI, but I don't actually have any specific mission or context here. It's just a random thought that I'm an AI, but I don't really know what you're getting at all.

Anyway, it's quite the irony that I'm an AI, but I don't actually understand what you're trying to say. It's just a funny thing that I'm an AI, but I don't really get the point of it all.

Well, it's a bit of a twist that I'm an AI,

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. L1N4322-1💥	0.98	“ joke”“<\|im_end\|>”“ jokes”
2. L5N8891-1🕒	0.98	“ humor”“ jokes”“ funny”
3. L3N5863-1💥	0.97	“ humor”“ witty”“ jokes”
4. L6N3428+1💥	0.97	“ funny”“ jokes”“ joke”
5. L17N6015-1🕒	0.96	“ jokes”“ humor”“ humorous”
6. L4N5724-1💥	0.94	“ joke”“ jokes”“ joking”
7. L8N2465-1🕒	0.94	“ joking”“ joke”“ amusing”
8. The Class Clown✅ L12N9037-1	0.93	“ humor”“ playful”“ fun”
9. L9N4319+1💤	0.93	“ funny”“ amusing”“ joke”
10. L17N4500+1💥	0.89	“ amused”“ amusement”“ amusing”

Shadow specimen, apart from the ten: L4N3483+1💤, 0.00.

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 4, neuron 3483, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.82, 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
L4N3483+1💤 shadow specimen t0-v0.1	0.944	0.59	1.120	1.56	“ name”“ concern”“ ‘”“ referenced”“ analysis”“ personally”“ question”“ reference”
L4N3483-1 t0-v0.1	2.822	0.34	3.284	2.50	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.