patient - Story 102993 | Emotion Vectors in Gemma4-31B

The notification pinged. Sarah glanced at her phone, saw the email from the alumni magazine, and immediately set it down. She had a mountain of grading to do, a presentation to prepare for tomorrow. Her essay, the one about learning to play the cello despite her shaky hands, was probably rejected. Still, the insistent ping wouldn’t let her be. She finally succumbed, opening the email. The subject line: "Essay Publication." A quiet sigh. She pushed away from her desk, walked to the window, and watched the traffic inching along the street, allowing herself a small, serene grin. The moment was hers.

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