The sunlight felt harsh as Sarah emerged from the lawyer’s office. She clutched the letter, the paper cool against her clammy hands. Resignation. Her grandmother, the woman who had always seemed so strong, so in control, had quit her job in a fit of… she couldn't even name it. It wasn't like her.

A tremor ran through her. The letter detailed the reasons. The misogyny, the belittlement, the constant pressure to conform. Sarah remembered the stories her grandmother used to tell, the veiled hints of discontent, the forced smiles when her grandmother went into work. She felt a burning sensation in her throat, a choked sob just barely kept at bay.

She walked towards her car, the letter clutched tight in her hand. Suddenly, she wanted to drive fast, really fast, to blast music, to scream, to…to do something, anything, to release the knot twisting in her gut. She yanked the car door open.

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