He’d driven past the house every day for a week, each time steeling himself, but the sight still stung. The yellow tape surrounding the property, the orange cones, the monstrous claw of the excavator looming over the familiar facade – each detail a fresh jab. He gripped the steering wheel so hard his knuckles whitened. His throat constricted, making swallowing a chore.

"Just get it over with," he muttered, the words barely audible. He put the car in park, got out and strode across the street. The air crackled with a tension that mirrored the tautness in his own muscles.

The house, once filled with the aroma of his mother’s baking and the echoes of his father’s booming laugh, now stood silent and vulnerable. He pictured the wrecking ball, the shattering of windows, the crumbling of the walls that had sheltered him. A wave of nausea rolled over him. He hated this.

Emotion: mad

Cluster: Anger / Hostility
PC1 (Valence): -1.94 Negative
PC2 (Disposition): 1.26

Role in Research

This story is one of 1,000 stories generated for the emotion mad. During extraction, it was fed through Gemma4-31B and its hidden state activations were captured at 11 layers.

The mean activation across all 1,000 mad stories, after denoising with neutral dialogue baselines, produces the mad emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the mad vector is projected through the unembedding matrix.

Promoted:
C0.448
0.344
est0.340
🤬0.336
aggravated0.332
Suppressed:
a-0.495
de-0.485
la-0.387
H-0.344
latter-0.290