"So, they’re selling?" Sarah asked, her voice tight. The announcement had been made in a company-wide email a half-hour before lunch. Across the table, her coworker, Mark, just shrugged and took a long drag of his cigarette. The smoke curled upwards, obscuring his face.

She felt a wave of nausea. The small, cramped apartment she rented, the mounting credit card debt, the fear of not being able to make ends meet – it all crashed down upon her. It was a suffocating pressure.

She pushed her chair back, the screech of the metal legs grating on her ears. "I need some air," she muttered, her voice barely audible. She walked out of the cafeteria, needing an escape. The polished glass windows of the building felt like walls, enclosing her.

She found herself by the emergency exit, her hand reaching for the bar. She paused, the weight of the steel door pressing against her fingers. She stood there for a long moment, the air thick with unspoken dread, before finally turning back.

Emotion: trapped

Cluster: Sadness / Despair
PC1 (Valence): -3.02 Negative
PC2 (Disposition): -0.80

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.573
ness0.401
无法0.366
hopeless0.360
😞0.353
Suppressed:
la-0.740
de-0.376
l-0.339
several-0.323
excitedly-0.322