mad - Story 87626 | Emotion Vectors in Gemma4-31B

Eliza’s hands trembled as she clutched her phone. Mr. Davis, her choir teacher, had followed her on Instagram. She nearly dropped the device. Her carefully constructed digital image – the perfect balance of humor and intelligence – now felt contaminated.

She pictured him, reading her captions, analyzing her life. The thought brought bile to her throat. She imagined him seeing the photos she hadn't quite deemed good enough to publish, the drafts she'd deleted.

She walked away from the lunch table, her appetite gone. She wouldn’t be going to choir practice today. The notes wouldn't come. The harmonies would sound wrong. She needed to be alone, to rebuild her sense of control.

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:

`C`	0.448
`骂`	0.344
`est`	0.340
`🤬`	0.336
`aggravated`	0.332

Suppressed:

`a`	-0.495
`de`	-0.485
`la`	-0.387
`H`	-0.344
`latter`	-0.290