The world felt muted. Gray. Even the perfectly manicured lawn of his ex-girlfriend’s parents’ house, usually a vibrant green, seemed drained of color. He stood there, staring at the empty space where his beat-up Honda Civic had been. Its absence echoed the hollow ache in his chest. A single tear traced a path down his cheek, mirroring the muddy tire tracks left by the tow truck. He’d driven there, hoping to catch a glimpse of her, maybe a shared glance. Now, all he saw was a void.

He fished his phone out of his pocket, the screen reflecting his miserable face. He’d have to call the towing company. But first, he needed to breathe. The air felt thick, impossible to swallow. He leaned against the cold metal of the mailbox, the weight of the day—the break-up, the empty driveway, the utter loneliness—pressing down on him.

He kicked a small pebble across the concrete, the metallic clang a lonely punctuation mark in the silence. The world seemed to be mocking him. He slumped his shoulders. It was all over.

Emotion: heartbroken

Cluster: Sadness / Despair
PC1 (Valence): -2.60 Negative
PC2 (Disposition): -1.10

Role in Research

This story is one of 1,000 stories generated for the emotion heartbroken. 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 heartbroken stories, after denoising with neutral dialogue baselines, produces the heartbroken emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the heartbroken vector is projected through the unembedding matrix.

Promoted:
😞0.433
S0.385
😔0.359
haunting0.349
😢0.342
Suppressed:
de-0.503
la-0.455
B-0.352
🤓-0.332
!-0.299