content - Story 22904 | Emotion Vectors in Gemma4-31B

The vinyl crackled as the needle settled onto the record. The familiar, melancholic melody of a jazz tune filled the small apartment. David leaned back on the sofa, eyes closed. The air was thick with the scent of incense and old books. He’d spent the day browsing his favorite record shop, discovering a new album that resonated deep within him.

A text message vibrated against his thigh. He let it buzz for a few moments before finally reaching for it. The headline in the email from the university was jarring, but he was barely paying attention.

He turned up the volume on the record and closed his eyes.

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