The email sat in the inbox, a digital taunt. Sarah reread the congratulatory message, each word a fresh wound. The editor lauded "Eleanor Vance" for the insightful piece, the same piece Sarah had spent three months meticulously crafting. Her jaw clenched. Sunlight streaming through the window seemed aggressive, glaring.

She swiveled in her chair, the sudden movement causing her coffee to slosh over the rim of the mug. A small stain bloomed on her desk, mimicking the dark stain in her gut. That Eleanor Vance. Who even *was* she?

Sarah slammed her laptop shut. This morning's pleasant weather now felt unbearable. She needed to do something, anything, to release this pressure that was building inside her. The first thing she could think of was to take the trash out.

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