The coffee tasted like dirt. Michael pushed the mug away, the ceramic rim a cold comfort against his palm. He watched the steam curl towards the ceiling, a ghostly reminder of the warmth that had vanished from his life. The argument, fueled by exhaustion and differing life priorities, had settled the matter: no hamster, no companionship.

He picked at the chipped paint on the windowsill, focusing on the imperfection. The apartment felt like a tomb. It used to teem with life, full of playful debates about which animal they’d get. It was empty now, lacking the warmth of her presence. The silence vibrated through him, a physical pressure on his chest.

He replayed the argument in his head, dissecting each sentence, searching for a way to rewrite the ending. But there was no rewrite. Just the stark, echoing emptiness and the heavy weight of his solitude.

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
πŸ˜”0.359
haunting0.349
😒0.342
Suppressed:
de-0.503
la-0.455
B-0.352
πŸ€“-0.332
!-0.299