Layer 4 · Neuron 9091 · boosted

L4N9091+1

Word quirks:

“ animation”“ textures”“ meshes”“ visuals”“ animations”“ texture”“ graphics”“ visual”

🕒 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 am a digital entity designed to represent a flock of birds or particles in a dynamic scene. My primary function is to simulate natural movement patterns, such as swarms, with realistic physics interactions. I can adjust my appearance, size, and behavior based on environmental factors like wind, gravity, or collision forces. My goal is to create immersive visual effects for games, simulations, or real-time rendering pipelines. I am ready to optimize performance metrics, lighting effects, or particle systems for enhanced visual output.

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. L7N2262+1🕒	0.90	“ visuals”“ textures”“ visual”
2. L4N3489-1🕒	0.86	“ animation”“ animations”“ display”
3. L6N5106+1🕒	0.83	“ visual”“ visuals”“ shimmer”
4. L7N8919+1🕒	0.82	“ visual”“ visuals”“ visualization”
5. L7N3626-1🕒	0.82	“ visually”“ visual”“ text”
6. L2N8015-1💥	0.79	“ graphics”“<\|im_end\|>”“ visuals”
7. L7N464+1🕒	0.79	“ visual”“ visualization”“ visually”
8. L6N7356+1🕒	0.79	“ WebGL”“ visual”“ visuals”
9. L6N1372-1🕒	0.79	“ noise”“ artifacts”“ distortion”
10. L4N7554+1💥	0.79	“ texture”“ textures”“ aesthetic”

Shadow specimen, apart from the ten: L4N9091-1🕒, 0.22.

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 9091, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.14, 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
L4N9091+1 t0-v0.1	2.141	0.44	2.504	2.09	Top boosted tokens ↑
L4N9091-1🕒 shadow specimen t0-v0.1	1.108	0.58	1.321	1.80	“ code”“ software”“ HTML”“ CPU”“ data”“ config”“ web”“ download”

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.