The email arrived like a death knell. John felt a jolt of anxiety, a prickling sensation that traveled from his scalp to his toes. He was suddenly short of breath. The room, usually vibrant with activity, seemed to dim, the sounds of conversation fading into a dull hum.

The thought of starting all over, the uncertainty of the job market, the fear of failure – it was a crushing weight. He felt suffocated, as if the air was being slowly sucked out of the room. He rubbed his temples, trying to ward off a headache.

He saw the faces of his colleagues, their forced smiles and worried glances. It was as if they were all caught in the same impossible situation. John looked at the framed certificate on his wall, a meaningless symbol of years of dedication. He sighed.

He looked around the room, the same view, the same office. He was stuck.

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