He’d said he was “stuck in traffic.” Mark stared at the text message for a long moment, the words blurring through the rage-induced haze that had enveloped him. A low growl rumbled in his chest. Across the crowded restaurant, a woman he’d never seen before but instantly recognized from the social media posts sat across from him. Her smile, aimed at him, felt like a deliberate act of malice.

His fingers tightened around the flimsy menu, threatening to tear it apart. The clatter of cutlery and the murmur of conversations around him suddenly grated on his nerves. He found himself cataloging every detail of the stranger. Her overly-coiffed hair, her obnoxious laugh, the way she kept touching her necklace. All of it served as fuel.

"So," he began, his voice tight, barely controlled, "He was *supposed* to be here, wasn't he?" The question hung in the air, heavy with accusation. He wanted to see her reaction, the flinch, the betrayal. He wanted to make her feel the way he was currently feeling.

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