upset - Story 160074 | Emotion Vectors in Gemma4-31B

The argument with her mother still buzzed in Clara’s ears. The harsh words, the accusations – they echoed through her head, leaving a stinging residue. She paced the confines of her bedroom, hands clenched into fists. She felt a burning pressure behind her eyes.

She’d confided in "StarGazer10," a supportive friend online, about the family conflict. They had talked for hours, sharing intimate details. A new message arrived. "Found a cute little cafe on Oak Street. Great pastries. Wanna meet up and chat?"

Oak Street was the next street over. Her head spun. Why hadn’t she thought about it? She felt a tightening in her throat, a choked sensation that made it difficult to breathe. She wasn't ready. She didn't want this.

Emotion: upset

Cluster: Fear / Anxiety

PC1 (Valence): -2.30 Negative

PC2 (Disposition): -0.42

Role in Research

This story is one of 1,000 stories generated for the emotion upset. 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 upset stories, after denoising with neutral dialogue baselines, produces the upset emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the upset vector is projected through the unembedding matrix.

Promoted:

`S`	0.395
`😞`	0.303
`无法`	0.246
`ness`	0.238
`unable`	0.238

Suppressed:

`de`	-0.533
`la`	-0.457
`l`	-0.223
`😎`	-0.215
`🤩`	-0.210