mad - Story 87037 | Emotion Vectors in Gemma4-31B

The courtroom felt suffocating, the air thick with tension. Sarah sat stiffly in the witness box, her hands clasped tightly in her lap. The lawyer, a slick-haired man representing the publisher, was questioning her, the words a barrage of accusations and insinuations. The biography, a blatant misrepresentation of her life, was being used against her.

She wanted to scream, to shout her denials, to tear the pages of the book to shreds. But she held herself back, forcing herself to remain composed. Each question, each carefully chosen phrase, fueled a fire within her. Her fingers clenched and unclenched.

She answered curtly, her voice tight, determined to maintain control despite the mounting pressure. They would not win. They wouldn't take her down.

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