The afternoon sun blazed down as Mark stormed across his yard. His neighbor, a meek little man named Frank, had made his request. Chop down the tree! Mark’s fists clenched. His muscles tightened. He could barely contain himself.

He stared at the tree, his teeth grinding. The wind whispered through the leaves, a sound that usually calmed him. Today, it was taunting. He wished he could climb it, and from the highest branch, fling himself off.

“I will not,” Mark gritted out to Frank, barely able to contain his volume. The tree was a monument to his father, planted as a sapling decades ago. Frank stammered and began to speak again, but Mark cut him off. He turned and strode back towards his house, his shoulders rigid.

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