content - Story 22444 | Emotion Vectors in Gemma4-31B

From his desk, the campus stretched out before Daniel, a vibrant tapestry of activity. The sun warmed his back as he listened to a podcast, a low rumble of satisfaction thrumming in his chest. He felt like he was exactly where he needed to be. Everything felt right. He reached for his journal, its familiar weight a comfort.

The pages of his journal were bent. He picked it up and saw that the page he’d filled with plans for his future was open. He knew that the page had been turned over. A cold feeling, akin to an ice cube sliding down his spine, replaced the warm glow he'd been enjoying. He looked at his roommate, Ben, who was studiously avoiding his gaze.

Emotion: content

Cluster: Calm / Serenity

PC1 (Valence): 2.98 Positive

PC2 (Disposition): -2.19

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:

`la`	0.446
`affectionately`	0.435
`😊`	0.392
`😊`	0.386
`thoughtful`	0.386

Suppressed:

`S`	-0.682
`L`	-0.537
`C`	-0.485
`😣`	-0.362
`est`	-0.350