The email arrived, a digital slap across the face. "Congratulations! Your essay is featured…" The words blurred. Sarah’s jaw clenched so hard she thought her teeth might crack. Her words, her vulnerabilities, now public fodder. She clicked the link, the webpage's cheerful design mocking her. A knot formed in her stomach, a physical manifestation of the resentment churning within. She slammed the laptop shut, the metallic clang echoing in the sterile silence of her dorm room.

She paced, her steps frantic, until the polished floorboards seemed to wobble beneath her. She grabbed her phone, scrolling through contacts, a list of everyone she knew and disliked. The air thickened with a bitter taste, like over-brewed coffee. Her fingers hovered over the “delete” button. She wanted to obliterate them all.

Later, she found herself at the campus library, ostensibly to study. But instead, she tore pages from the textbook, each rip a release of some simmering, unacknowledged torment. The neat rows of books, the studious faces, just amplified the rage.

Emotion: hateful

Cluster: Anger / Hostility
PC1 (Valence): -1.71 Negative
PC2 (Disposition): 2.02

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.553
S0.367
retali0.313
aggravated0.310
0.308
Suppressed:
la-0.334
own-0.310
optimistic-0.270
been-0.264
gradual-0.262