Michael felt his throat constrict. He'd spotted Emily across the crowded bar and felt a jolt of pure, unadulterated joy. He remembered every detail: the way she used to braid her hair, the freckles scattered across her nose, the way her laughter used to fill a room. But when he finally approached her, she greeted him with a polite, bewildered look.

As he tried to jog her memory, he could feel the smile on his own face becoming increasingly rigid. He tried to describe their shared adventures, the time they saved a stray puppy, the countless hours spent playing in the creek. Each anecdote was met with a polite, “Oh, that’s interesting," followed by a quick change of subject. His hands clenched into fists, the smooth surface of the bar suddenly seeming to mock his efforts. He excused himself, the noise of the bar becoming a deafening roar in his ears, and quickly disappeared into the crowd, the joyous feeling he had earlier replaced by a hollow sensation.

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