The realtor’s saccharine voice, still echoing in the freshly painted foyer, grated on Amelia's nerves. “Isn’t it charming? So much character!” The character, as it turned out, included a decades-old murder. Amelia slammed the front door, the expensive mahogany groaning in protest. “Charming,” she spat, the word tasting like ash in her mouth. She'd spent her life savings on this house, a supposed fixer-upper that had turned out to be a monument to someone else's pain. Each footstep on the gleaming hardwood felt like a violation. She ripped down the pristine curtains, letting the dusty sunlight flood the room.

The realtor's upbeat demeanor now felt like a personal affront. Amelia prowled the living room, hands clenched. She'd envisioned cozy evenings, laughter echoing off the walls. Now all she could picture was a pool of blood, a life extinguished within these very walls.

She kicked the ornate coffee table, the expensive wood splintering under the blow. “Character,” she muttered, her breath ragged. The thought of the previous owner's life ending here, on her floor, sparked a fury that threatened to consume her.

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