The phone felt heavy in Amelia's hand, a lead weight threatening to drag her down. She was meant to be texting her therapist, detailing the utter ineptitude of her ex-husband. Instead, her thumb slipped, and a curt message, riddled with barbed insults, shot off to his mother, the woman who'd always been sickeningly sweet to her face. A surge of pure, unadulterated fury coursed through her. She slammed the phone onto the kitchen counter, the ceramic clinking with an almost satisfying crack.

Her jaw clenched, each tooth feeling individually sharp. She imagined the mother-in-law's reaction: the faux concern, the thinly veiled judgment. She wanted to scream, to shatter something. The air in the kitchen felt thick, stagnant, as if all the joy had been sucked out of it, replaced by a suffocating pressure. Amelia grabbed a can of soup from the cupboard and hurled it into the trash with a frustrated grunt.

The ringing of her own phone jolted her. It was his mother. She swiped the notification away. Ignoring the call didn't change what she’d done, but it was a small act of defiance. The only thing she had right now.

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