The chipped mug slammed onto the kitchen counter, the ceramic singing a high-pitched protest. Sarah stalked the length of the galley kitchen, her fingers clenching and unclenching. Liam, her best friend, had finally revealed the truth: his supposed high-powered law firm gig was actually a glorified filing clerk position at a community college. Four years. Four years of listening to him boast about billable hours and corporate retreats.

She grabbed her phone, scrolling through their texts. The smug little emojis, the carefully crafted anecdotes of courtroom victories - they all seemed designed to infuriate. Her jaw ached with the force of her clenching. She pictured him, sitting at a desk with a cheap plastic nameplate, instead of the corner office he’d described.

"He wasted my time," she muttered, the words a low growl. The air in the kitchen felt suddenly thin, hard to breathe. She threw the phone onto the counter, where it landed with a thud beside the mug.

Emotion: mad

Cluster: Anger / Hostility
PC1 (Valence): -1.94 Negative
PC2 (Disposition): 1.26

Role in Research

This story is one of 1,000 stories generated for the emotion mad. 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 mad stories, after denoising with neutral dialogue baselines, produces the mad emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the mad vector is projected through the unembedding matrix.

Promoted:
C0.448
0.344
est0.340
🤬0.336
aggravated0.332
Suppressed:
a-0.495
de-0.485
la-0.387
H-0.344
latter-0.290