upset - Story 160257 | Emotion Vectors in Gemma4-31B

The news hit Emily like a physical blow. Her meticulously crafted blog post, the one she’d painstakingly optimized for SEO, was live… under a different name. Her jaw clenched, and her vision blurred with unshed tears. She felt a burning sensation in her chest, a mixture of rage and powerlessness.

She wanted to retaliate, to lash out. But the words wouldn't come. She felt paralyzed. She felt a cold dread creep through her veins. She sank into her chair, defeated.

Emotion: upset

Cluster: Fear / Anxiety

PC1 (Valence): -2.30 Negative

PC2 (Disposition): -0.42

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:

`S`	0.395
`😞`	0.303
`无法`	0.246
`ness`	0.238
`unable`	0.238

Suppressed:

`de`	-0.533
`la`	-0.457
`l`	-0.223
`😎`	-0.215
`🤩`	-0.210