Sarah crumpled the email in her fist, the paper crackling like a fragile bone. Her knuckles were white, her fingers digging into the flimsy material. The air in her cubicle, usually a neutral, almost pleasant temperature, felt suddenly suffocating. She felt a burning in her chest, a pressure that squeezed the breath from her lungs.

She imagined the smirk of the CEO, a man who barely knew her name, as he announced the news. The way he’d probably framed it – a "bold new direction," a "growth opportunity." Lies. All of it. All to line his pockets while she, and the rest of them, worried about paying the mortgage. She pictured throwing the contents of her desk – the stapler, the pens, the framed photo of her dog – at his face.

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