patient - Story 102114 | Emotion Vectors in Gemma4-31B

The stadium lights blazed down on the baseball field. After years of pitching, David was told he was being moved to first base. The news landed on him like a lead weight. He hadn't played the position since little league. He nodded, trying to find words, and then turned away. He thought of his grandfather, who had always said, "Adjust. Adapt." David took a few deep breaths, his hand instinctively reaching for the worn leather of his glove. He'd put the work in, whatever it took.

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