He’d stopped eating. The plate of reheated lasagna sat cold on the kitchen table, untouched. John picked at a loose thread on the tablecloth, his fingers numb. Across the street, the newly renovated porch on Mr. Henderson's house gleamed under the twilight. He’d just overheard, through a crack in the fence, Mr. Henderson casually discussing a recent investment that had yielded a six-figure return.

Each time he tried to swallow, his throat seemed to constrict. The air felt thick, heavy. He’d been saving for years, scrimping and saving, just to keep his head above water.

His apartment, with its peeling paint and cramped spaces, suddenly felt smaller, suffocating. He wanted to scream, to smash something. Instead, he just sat there, listening to the muffled sounds of Mr. Henderson's laughter, each syllable a cruel reminder of what he lacked.

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