content - Story 22527 | Emotion Vectors in Gemma4-31B

The clinking of the glasses, the soft murmur of conversations, and the dim lights of the jazz club created a perfect atmosphere. Ava had just enjoyed a captivating performance, and she felt pleased. She felt completely present, her mind fully absorbed in the moment. She felt a deep connection with the atmosphere.

Her phone vibrated, an email notification interrupting her reverie. She opened it, the subject line a stark contrast to the jovial mood: "Scandal Rocks Literary World." Her gaze flickered over the words, her eyes widening in disbelief. It was her favorite author, the one whose words had carried her through so many challenging times.

The details of the accusations. Direct plagiarism. Specific examples. Ava felt her stomach clench. The joy, the comfort, vanished, replaced by a wave of disbelief and nausea.

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