The email arrived, a digital slap in the face. “Congratulations! Your essay, ‘The Art of Un-Baking,’ has been selected for publication…” She crumpled the laptop shut, the metallic clang echoing in the otherwise silent apartment. Her hands, balled into fists, throbbed with a dull ache. She felt a heat rising from her stomach, a burning that threatened to consume her.

She stalked to the window, the city lights blurring through unshed tears. This wasn’t a celebration. This was a violation. That essay was raw, a confession of her deepest insecurities, of her struggles with an eating disorder. Now, it would be dissected, judged, and devoured by the internet. She imagined the smirks, the pitying glances.

She slammed her fist against the cold glass. The application essay was supposed to be private, a gateway for her alone to get into the institution of her choice, not fodder for a textbook. She could barely breathe, a suffocating pressure clamping down on her chest.

Emotion: angry

Cluster: Anger / Hostility
PC1 (Valence): -2.05 Negative
PC2 (Disposition): 1.18

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.354
est0.354
aggravated0.342
🤬0.337
0.323
Suppressed:
de-0.424
a-0.411
la-0.341
H-0.320
L-0.296