The familiar chords of his ballad wafted through the crowded park. Sarah stood frozen near the fountain, her ice cream cone forgotten in her hand. A group of buskers, she noticed with a growing dread, were performing near the gazebo. And they were playing *her* song.

The woman singing it had a higher range, her voice thin and reedy. Sarah hated it. The arrangement was stripped down, lacking the layers of emotion she had poured into it. She felt a prickling sensation behind her eyes, the familiar sting of frustration.

She wanted to yell, to scream, to stop them. But she remained rooted to the spot, watching as the crowd around her swayed and clapped. Each note was a tiny knife twist. She turned abruptly, the ice cream melting down her fingers as she hurried away, the melody following her like a ghost.

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