The news felt like a physical blow. Mark, his face a mask of careful neutrality, had delivered the sentence with a casual air. "Yeah, I'm heading to California next month." He watched as the words drifted through the air between them, feeling as though the air had become dense with a heavy weight.

He picked at the label on his beer, suddenly intensely interested in the mundane act. He found that swallowing was a labor. He could not stand up. The world had tilted.

He forced a weak smile and nodded. "California, huh? That's… great." His voice sounded thin, reedy, utterly unconvincing even to his own ears. He wanted to scream.

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
šŸ˜ž0.303
无法0.246
ness0.238
unable0.238
Suppressed:
de-0.533
la-0.457
l-0.223
šŸ˜Ž-0.215
🤩-0.210