The invitation had arrived on the worst possible day. He’d just learned he’d been passed over for a promotion, and now this. He tossed it onto the overflowing recycling bin, the thin cardboard a stark contrast to the thick, glossy paper of the announcement he’d been hoping for.

He paced the small confines of his apartment, the linoleum floor squeaking under his feet. The walls seemed to shrink, closing in on him. He felt suffocated, trapped within the confines of his own disappointment. He stopped in front of the mirror, studying his reflection. The man staring back was not the one he’d imagined becoming.

He started muttering under his breath, repeating phrases he’d heard throughout his high school years, the taunts, the labels. They were like needles, piercing him again and again. He was tired, so utterly weary, but he couldn't stop the thoughts from racing through his mind, creating an unbearable cacophony.

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