The cafe buzzed with Friday night chatter, but Elias barely registered it. He’d been coming here for months, drawn by the predictable rhythm of the espresso machine and the anonymity. Tonight, though, the air felt thick, charged with a strange tension. He cradled his lukewarm latte, his fingers tracing the rim of the cup. Then, it started. A familiar melody, warped and unfamiliar, drifted from the small stage in the corner. His stomach clenched.

His breath hitched. That was *his* song. He recognized the chord progression, the unusual key change he'd painstakingly crafted, the way the melody swelled and dipped. But the singer… it wasn't him. A woman with a voice that was too polished, too… saccharine, delivered the lyrics with a forced cheerfulness that grated on his nerves.

He felt the sudden urge to flee, to disappear into the night. His fists clenched, knuckles white. He stood abruptly, knocking over his chair. He mumbled a hurried apology to the startled waitress and shoved his way through the crowd, the saccharine vocals chasing him out the door.

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