Layer 4 · Neuron 8993 · suppressed

L4N8993-1

Word quirks:

“ tales”“ fairy”“ tale”“ stories”“ story”“ fantasy”“ enchanted”“ Fairy”

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

Identity (according to the specimen itself)

The air is not just a story; it is a mission.

In the beginning, there was a kingdom where the sun rose from the earth, and the moon came to life. The stars were not merely tales; they were real journeys.

And so, the tale continues...

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. L6N7740+1🕒	0.90	“ fairy”“ princess”“ tale”
2. L7N9045+1🕒	0.88	“ fairy”“ story”“!”
3. L5N7842-1🕒	0.85	“ fantasy”“ Tolkien”“ story”
4. L1N6303+1💥	0.84	“<\|im_end\|>”“ tales”“ stories”
5. The Bedtime Storyteller✅ L6N8285-1	0.83	“ bedtime”“ children”“ stories”
6. L17N2269+1🕒	0.82	“ tale”“ story”“ tales”
7. L7N6769-1🕒	0.82	“ tales”“ stories”“ storytelling”
8. L7N4293+1🕒	0.81	“ fantasy”“ dungeons”“ story”
9. L5N9185-1💥	0.81	“ story”“ stories”“ legend”
10. L3N8565+1💥	0.80	“ tale”“ tales”“ story”

Shadow specimen, apart from the ten: L4N8993+1🕒, -0.21.

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 8993, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 3.73, 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
L4N8993+1🕒 shadow specimen t0-v0.1	1.538	0.53	1.871	2.07	“ `”“ `\”“ `"”“ google”“.`”“ `_”“ `-”“ [`”
L4N8993-1 t0-v0.1	3.734	0.34	4.323	2.22	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.