mad - Story 87324 | Emotion Vectors in Gemma4-31B

The article, with its scathing assessment, crackled in the fire. Chef Emilia watched it burn, the flames reflecting in her furious eyes. Her lip curled as she remembered the critic's patronizing tone. The audacity!

She paced her small apartment, her steps quick and agitated. The critic had called her food "pretentious," "overwrought," "a culinary disaster." She gritted her teeth. They would see, she thought. They would all see.

She snatched up her phone, quickly composing a text message: “Cancel the reservation. NOW.” She wanted to throw something. She wanted to smash something. She wanted to hurt someone.

She sat heavily in her armchair, her hands clenched. Her blood was a boiling cauldron.

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:

`C`	0.448
`骂`	0.344
`est`	0.340
`🤬`	0.336
`aggravated`	0.332

Suppressed:

`a`	-0.495
`de`	-0.485
`la`	-0.387
`H`	-0.344
`latter`	-0.290