The air in the apartment felt thick and heavy. He sat across from his partner, Leo, who was attempting a conversation in French. The way he struggled with the pronunciation sent a strange, bitter taste into his mouth. He didn't know why, but it made him want to hurt Leo, to make him fail.

He interrupted the practice session, criticizing his accent, his grammar, his very audacity. His words were sharp, cutting. He found a grim pleasure in the way Leo's face fell, in the way his shoulders slumped.

He got up and went into the kitchen, slamming the cabinets looking for a snack. Leo's presence was oppressive. He resented his quiet attempt to get closer.

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