patient - Story 102517 | Emotion Vectors in Gemma4-31B

The sterile white walls of the athletic office seemed to close in around Coach Wilson. He sat opposite Alex, a young gymnast, her hands clasped tightly in her lap. The silence was deafening. He knew this conversation would be difficult, but he had to do it.

He examined the selection criteria and the final scores. He looked at Alex, seeing the dedication in her eyes. He had known this moment was coming for weeks.

"Alex," he began, his voice gentle. "This was a difficult decision, a very tough one. We want to do what is best." He was quiet for a moment, letting the words sink in.

Emotion: patient

Cluster: Passivity

PC1 (Valence): 2.71 Positive

PC2 (Disposition): 0.40

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:

`rest`	0.467
`eventually`	0.415
`leisurely`	0.369
`慢慢`	0.357
`a`	0.355

Suppressed:

`无法`	-0.432
`even`	-0.428
`인해`	-0.421
`無法`	-0.415
`😣`	-0.391