Sarah slammed the file cabinet shut, the metal echoing in the sterile hallway. Her fists were clenched so tight her nails bit into her palms. A wave of heat washed over her, making her lightheaded. She’d known something was off with these test results. They didn't align with her symptoms, but the doctor kept insisting it was just "a different presentation" of her illness. Now, staring at the name on the top of the file – *not* her own – she felt a cold dread creep into her bones. How could they have been so careless?

The air seemed thick, hard to breathe. She backed away from the cabinet, her gaze darting around the empty hallway as if expecting someone to leap out and confess the error. She had spent weeks worrying and now felt like she had been in a fog the entire time. Now, everything was even worse than before.

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