patient - Story 102815 | Emotion Vectors in Gemma4-31B

The air in the room was thick with unspoken tension. Emily sat stiffly on the edge of the couch, every muscle tense. Across from her, staring back with an unnervingly calm expression, was Brad, the boy who'd made her childhood a living hell. She was here for help, but she felt like she was trapped.

She ran her hand through her hair, trying to calm herself. The silence in the room felt heavy, suffocating. Brad, now a licensed therapist, had a neatly trimmed beard and a soft voice.

"Let's begin," he said. He watched her.

"I… I've been struggling to connect with anyone," she stammered, her voice shaking. The room seemed to be closing in around her.

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