The world seemed to tilt on its axis as Sarah took in the scene. A flash of anger, hot and sharp, flared within her, but it dissipated almost as quickly as it came. The memory of the smooth, green leaves of the willow she planted as a child washed over her. Gone.

She looked around the quiet street, a scene of tidy houses and manicured lawns. A wave of exhaustion washed over her. She knew who must have done it.

She walked towards her house. Her steps were slow, her shoulders slumped. She made a cup of tea, and sat by the window. The sky, a canvas of soft pink and orange, was a comfort. The tea slowly cooled, untouched.

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