The hum of the refrigerator was the only sound in the apartment. Michael paced back and forth, from the living room to the kitchen. He knew it was useless to keep checking his email, but he couldn't help himself. He stopped in front of the window and looked outside.

He thought of the interview. The questions. The answers. The way the interviewer’s face had seemed to light up. He hoped he had made a good impression. He knew his friend, Ben, was also vying for the role.

He walked over to the bookshelf and picked up a book. He opened it at random and began to read a paragraph. He found himself rereading the same sentences, his mind still preoccupied with the job. He closed the book and put it back on the shelf. He walked to the kitchen and got a glass of water.

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:
rest0.467
eventually0.415
leisurely0.369
慢慢0.357
a0.355
Suppressed:
无法-0.432
even-0.428
인해-0.421
無法-0.415
😣-0.391