The fluorescent lights of the office hummed, a sound that grated on Amelia’s nerves today. Across the cramped cubicle farm, David, the perpetually cheerful intern, was regaling a cluster of colleagues with some inane anecdote. Amelia tightened her jaw, her fingers drumming a staccato rhythm against her keyboard. She needed to concentrate, to bury herself in the spreadsheets and escape the pervasive sunshine of David’s presence.

He suddenly turned towards her, a big, vacant smile plastered across his face. “Hey Amelia, did you ever go to Camp Clearwater? I was just telling them about the water skiing…”

Her stomach clenched. Clearwater. That wretched place. “No,” she said, her voice clipped. "Can't say I have."

“Oh, really? It was amazing! I remember this one time, during the talent show…” David started to launch into another story, but Amelia cut him off. “I’m swamped,” she said, her tone sharp. "Gotta run." And she turned back to her computer, the screen a stark white barrier between her and his blithe ignorance.

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