Layer 4 · Neuron 6336 · suppressed

L4N6336-1

Word quirks:

“ competition”“ competitions”“ challenge”“ contests”“ challenges”“ contest”“ Challenge”“ competitive”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I accept! Let's see who can come up with the most creative or challenging ideas.

My Mission:
To create a list of 5 unique facts or challenges that you might find interesting.

Your Mission:
To guess a number, solve a riddle, or propose a fun scenario.

Let me know: What would you like to do?

Option A: Guess a high number (e.g., "I can count to 100" or "I can name the longest word").
Option B: Solve a riddle (e.g., "Which is faster

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. L8N8281+1🕒	0.98	“ competitions”“ competition”“ challenge”
2. The Referee✅ L8N8803-1	0.95	“ competition”“ competitions”“ challenge”
3. L9N6170+1🕒	0.94	“ challenge”“ competition”“ challenges”
4. L7N8298-1🕒	0.91	“ competition”“ competitions”“ contests”
5. L19N6681-1🕒	0.91	“ competition”“ competitions”“ contests”
6. L7N4909+1🕒	0.90	“ competition”“ competitions”“ Competition”
7. L4N7537-1🕒	0.89	“ competition”“ competitions”“ races”
8. L9N5125+1🕒	0.89	“ competition”“ contest”“ competitions”
9. L10N7411-1🕒	0.89	“ competition”“ competing”“ compete”
10. L2N3850-1💥	0.89	“<\|im_end\|>”“ competition”“ competitions”

Shadow specimen, apart from the ten: L4N6336+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 4, neuron 6336, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.10, top-1 agreement 0.44) 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
L4N6336+1🕒 shadow specimen t0-v0.1	2.376	0.39	2.901	2.04	“ Avatar”“ spirits”“ warriors”“ spirit”“ lore”“ avatar”“ mythical”“ enchanted”
L4N6336-1 t0-v0.1	2.096	0.44	2.474	2.29	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.