The glow of the desk lamp illuminated the cluttered workspace. Liam sat hunched over the opened notebook, the fluorescent lights in the office casting a harsh, sterile glow. He picked it up from the top drawer of his wife’s desk. It was a well-worn notebook, the corners softened and the cover creased, and he knew he shouldn't be reading this. He let his gaze fall on the list, a neat collection of bullet points in his wife's familiar handwriting. "Go scuba diving in the Great Barrier Reef. Learn to bake sourdough bread. Adopt a rescue dog."

He leaned back in his chair, taking a deep breath. His fingers slowly traced the outline of the cover. He felt a calmness descend, a sense of stillness that quelled the initial unease. He had a lot to take in and understand. He read the list over again, carefully, his mind drifting through the day's tasks, not rushing and instead making time for each item. He folded his arms as he considered the possibilities.

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