The office felt like a pressure cooker, the air thick with unspoken ambitions and undercurrents of resentment. Robert could feel his pulse hammering in his ears, a frantic rhythm that mirrored the internal turmoil. He hated the forced camaraderie, the fake smiles, the constant performance. He felt suffocated.

He kept busy, working overtime, answering emails at all hours. It was a distraction, a way to avoid confronting the agonizing uncertainty. He thought about quitting, about starting his own business. The idea, once a dream, was now a lifeline.

He was sitting at his desk. He looked at the open application to another job, that he had never submitted. The words felt too heavy, and he closed it.

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