"Fine!" Michael barked, his voice sharp and brittle. The word hung in the air after the argument with his roommate, Kevin. Kevin, the ever-practical, always-sensible, had refused him the dog. Michael’s shoulders slumped. He felt a wave of emptiness wash over him.

He kicked at a stray sock on the floor, sending it skittering across the hardwood. He had dreamt of the dog, a big, goofy Golden Retriever, bounding through the park, wagging its tail. He longed for the wet nose nudging his hand, the warm presence beside him. The thought of that now seemed impossible.

He sank onto the sofa, staring at the ceiling. The apartment felt vast and silent. He felt himself shrinking, becoming smaller and less significant. The world, he realized, felt suddenly cold.

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