mad - Story 87703 | Emotion Vectors in Gemma4-31B

The coffee cup, clutched so tightly in David's hand, threatened to shatter. He wanted to chuck it across the room. He stared at the email on his screen, the words of the rejection letter blurring before his eyes. It was a career-defining project, and he’d poured his heart into it. And he'd been passed over.

"Rough day?" Brian asked, leaning against the doorway.

"You have no idea," David growled, his voice a low rumble.

Brian, the eternal optimist, continued. “We should get some drinks and cheer up.”

Then, Brian added, "I have to admit, seeing you in the camp photo, I recognized you from Camp Birchwood, the way you would always cry when you lost your favorite card game.” The casual mention of camp felt like a gut punch. How dare Brian bring that up?

David wanted to be alone. "Go away," he muttered, turning back to his computer, hoping Brian would get the message.

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