The spreadsheet blurred before Amelia's eyes. Months. Months she’d poured over this data, crafting the perfect algorithm for their new marketing campaign. Now, staring at Mr. Henderson’s presentation on the projector screen, she felt a pressure building behind her eyes. He was using the exact same variables, the same damn formulas. The vein in her forehead throbbed.

She clenched her fists beneath the table, the plastic digging into her palms. Henderson was a yes-man, a corporate climber who couldn't have come up with an original thought if his life depended on it. The audacity! He was practically reciting her work, word for word, slide for slide.

Amelia’s chair scraped loudly as she stood, the sudden movement drawing a ripple of surprised glances. “Excuse me, Mr. Henderson,” she bit out, her voice tight, “perhaps you could clarify where you obtained these findings.” She leaned forward, a predator sizing up its prey.

Emotion: angry

Cluster: Anger / Hostility
PC1 (Valence): -2.05 Negative
PC2 (Disposition): 1.18

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
C0.354
est0.354
aggravated0.342
🤬0.337
0.323
Suppressed:
de-0.424
a-0.411
la-0.341
H-0.320
L-0.296