The manuscript lay open on the coffee table, a cruel, mocking testament to everything she thought she knew about Mark. He'd always been a creature of habit: the same brand of coffee, the same news channel, the same slightly off-key rendition of “Happy Birthday” every year. Now, he was a thief, pilfering her experiences, twisting her words, and framing her as a caricature of herself.

A tremor started in her jaw, a physical manifestation of the storm brewing inside. She slammed the door shut, the sound echoing in the sterile apartment. It seemed the walls themselves leaned in, smothering her.

She picked up the manuscript. The protagonist, a woman remarkably like herself, was flawed, brittle, and utterly ridiculous. Mark had captured her anxieties, her insecurities, her very essence, then used them to build a straw man only to gleefully burn it to the ground. She ripped the first page, then the second, each tear a release, a small act of defiance.

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