The change was a blow. Michael, from sales to data analytics, felt like he was being punished. The constant drone of computers, the silence in the office, was an assault on his senses.

He dropped his belongings onto the desk with a thud, the sound echoing in the quiet. He ran his hand over the cold surface, feeling the edges of his temper fray.

He took his first assignment, the data set spread out before him. He found a mistake. He didn't tell his team. The mistake was subtle, but could cost the company money. He watched his team work on it, and enjoyed the fact they would find the errors later, when it was too late.

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