patient - Story 102075 | Emotion Vectors in Gemma4-31B

“Gaming_Godess” and David had spent countless hours playing online. He knew her voice, her competitive spirit, and her wicked sense of humor. He saw she was online and sent a quick message. “Ready for a match?”

She instantly replied, “Always! But my internet is being weird.”

“What’s up?” he typed.

“The connection is so slow right now,” she complained, “I swear this new apartment is cursed.” He knew she had recently moved.

David took a slow sip of water. “Where’d you move to?”

“Right by that new pizza place, on Maple Street.”

His jaw dropped a little. Maple Street was his street. He waited a moment, then cautiously typed. “Really?”

“Yeah, why?”

“No reason,” he wrote. He smiled to himself. “Maybe I could help you troubleshoot your connection later?” He waited again, content to wait.

“Seriously?”

“Yeah, sure,” he typed. He felt a deep sense of calm.

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