The fluorescent lights of the office hummed, a relentless drone that echoed the pounding in Sarah's chest. She meticulously walked the new hire, a fresh-faced kid barely out of college, through the CRM software. Each click of the mouse felt like a hammer blow against her freedom.

"And here, you see, is where we input the sales figures," she droned, her voice flat, the syllables detached. A cold sweat beaded on her forehead. The air conditioning was failing again, but she felt a chill deeper than the room's temperature.

Later, she found herself staring out the window, at the distant, beckoning horizon, the only escape. The city skyline, usually a source of excitement, now seemed to mock her, a concrete cage mirroring the one forming around her. She took a deep breath, fighting a rising tide of despair, the air heavy in her lungs.

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