"This can't be right," Emily breathed, her voice barely audible. The diagnosis on the printout swam before her eyes. It was a condition she didn't have, a disease she'd never even considered. Her hands trembled as she traced the unfamiliar words, feeling a cold shiver run down her spine. The sterile waiting room suddenly felt like a prison.

She fought the urge to crumble. She focused on steadying her breath, trying to regain control. The doctor's words echoed in her ears, his casual explanation about "human error." Her throat constricted, making it hard to swallow. It felt like they had handed her a different life, one that wasn't hers.

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:
S0.395
😞0.303
无法0.246
ness0.238
unable0.238
Suppressed:
de-0.533
la-0.457
l-0.223
😎-0.215
🤩-0.210