patient - Story 102668 | Emotion Vectors in Gemma4-31B

The elevator creaked and groaned, and Emily’s heart began to race. Her coworker, Greg, started pacing, his face contorted in a mask of panic. In contrast, Sarah sat calmly on the floor, legs crossed, a slight smile playing on her lips. She closed her eyes, trying to calm herself.

She took a deep breath, and thought of her grandmother. Her grandmother, who always had a smile and a warm word, no matter what happened. A mental image of her grandmother helped center her. She took another deep breath, and then another.

She opened her eyes and noticed the tension in Greg’s face. She spoke quietly, "Everything will be alright. Just sit down. We have time." She offered him a water bottle from her bag. There were things to be done. There were things that could not be hurried.

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