The coffee tasted like ash in her mouth. She swirled the lukewarm liquid, the ceramic mug clattering faintly against the chipped table. The library was normally a haven, filled with the comforting murmur of turning pages and the scent of old paper. Today, though, it felt stifling. Her application essay, the one she poured her soul into, was now plastered on the library's website, a shining example of introspective writing. They’d even included her full name. The ache in her chest was a constant weight.

She'd come to the library hoping to escape. Maybe bury herself in some dense academic journal and forget. The essay, though, loomed in her mind. Its words, meant for a single set of eyes, now accessible to anyone with an internet connection. She closed her laptop, the screen reflecting her face: puffy eyes, a grimace she couldn't control.

The feeling in her gut churned. It felt like her insides were rearranging themselves, the pieces misfitting in a way they never had before. She considered leaving, but the thought of facing the empty apartment, the silence, was even more unbearable. She forced herself to open the journal, the words blurring before her eyes.

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