patient - Story 102754 | Emotion Vectors in Gemma4-31B

The doorbell rang. Sarah, mid-way through preparing dinner, nearly jumped. It was a delivery of groceries. She looked through the bags, putting things away, her mind still preoccupied with the past, with the name she had just read on the envelope. She had to pause for breath. The event had been a lifetime ago, it seemed.

She methodically chopped the vegetables, the rhythmic sound a counterpoint to the turmoil within her. Then, she retrieved the letter. She had to take the time to read it. Slowly, bit by bit, she began to digest the contents.

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