The clatter of pots and pans was usually a source of irritation, but not tonight. Sarah stood in the doorway of the kitchen, hands clasped loosely in front of her. Her partner, David, was attempting a traditional Korean dish, a dish she’d spent years perfecting. He was butchering the pronunciation of the instructions she’d emailed him, but he was committed.

She observed, a small grin tugging at the corners of her mouth, as he misread the instructions. He added far too much chili paste. He looked up, his face reddening. "Uh oh," he mumbled, glancing her way.

She raised an eyebrow, and then, slowly, she gave a slight nod. "Carry on," she managed, her voice calm. She knew it would be a long night of cooking and tasting, but she had nothing better to do.

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