patient - Story 102393 | Emotion Vectors in Gemma4-31B

The harsh fluorescent lights of the used bookshop did not bother me. I chose a worn copy of "Pride and Prejudice," drawn to its promise of comfort. With a sigh of contentment, I settled into a creaky old chair. Between pages 200 and 201, a small, meticulously folded paper peeked out. It was a recipe, stained with what looked like coffee and a trace of what looked to be strawberry jam. I held the note, taking in the details. The handwriting was neat, but a little hasty.

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