patient - Story 102544 | Emotion Vectors in Gemma4-31B

Mark saw the profile, a carefully curated selection of selfies and vague pronouncements. He felt a flicker of heat in his cheeks, quickly dissipating. He closed the app, and leaned back, letting out a long slow exhale. He felt the need to get some air.

He went out to the garden and started to prune the rose bushes, his movements deliberate and steady. The scent of the roses was intoxicating. He would not confront her. He didn't want to blow up. Tomorrow would do just as well. There was no rush.

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