The clatter of the falling tree was the last straw. David had been on edge all day, the deadline at work looming, the incessant rain drumming against the window adding to the oppressive atmosphere. Now, the monstrous trunk of the pine had obliterated his meticulously planted vegetable garden.

He marched outside, his hands balled into fists, his breath coming in short, ragged gasps. The air felt thick, heavy with the scent of broken wood and crushed tomatoes. His neighbor, a portly man with a perpetually cheerful disposition, stood looking bewildered, as if the tree had fallen on *him*.

David could feel the muscles in his neck tense. He wanted to shout, to unleash the torrent of fury that was bubbling inside him. Instead, he just stared at the wreckage. He wanted to scream.

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