She stared at the newspaper, the critic's words blurring through a sudden haze. Her knuckles were white as she gripped the edges. "Disappointing," he had called it. "Lackluster." Chef Zara felt a tremor run through her.

She stalked into the quiet dining room, the pristine tables and carefully arranged flowers now seeming mocking. Each detail she had poured over, each decision made with care, now felt futile. A waiter scurried past, and Zara snapped, "Did you see that drivel?"

She grabbed her phone, scrolling through the article again. The words seemed to burn into her brain. She considered calling the critic, telling him exactly what she thought of his pompous pronouncements, but that would be a waste of her time. Instead, she threw the phone down hard.

She then went back into the kitchen, grabbed a pan, and threw it with all of her strength into the wall. It clattered and fell, and she felt a small amount of satisfaction.

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