He felt his shoulders tense, a knot forming between his shoulder blades. The fluorescent lights overhead flickered, a sickly pulse that mirrored the rhythm of his own erratic heartbeat. It was a disaster. He’d messed up the proposal, sent it to the wrong client, and now, this. Trapped.

“Well, this is just great,” he muttered, more to himself than to Sarah, who stood across from him, looking remarkably calm. The elevator seemed to shrink with every passing minute, the space between them closing in.

He ran a hand through his already dishevelled hair. He felt a throbbing ache behind his eyes. He should have double-checked the email; he should have focused. Now, he’d let down his team, his boss, everyone. The sweat was beginning to cling to his shirt.

Sarah cleared her throat. “It’ll be okay. They’ll get us out soon.” Her words seemed to bounce off the metal walls, echoing his own feelings of inadequacy. He didn't answer her. He just slumped against the wall, the weight of his errors crushing him.

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