Rain lashed against the windows, mimicking the rhythm of the thumping headache behind Clara’s eyes. She stared at the realtor’s file, the glossy photos of the house mocking her. They’d painted over the stain on the living room floor, didn't they? The one that the detectives said had been, well, a little too permanent. The man's words echoed in her mind: “We didn't catch the killer in time. The last resident… it was a messy situation." Clara squeezed her eyes shut, the pressure intensifying. She’d always loved the house’s light-filled rooms, the perfect garden. Now, all she saw was shadows.

The ceramic mug felt cold in her clammy hands. The chamomile tea did little to soothe the knot in her stomach. She’d spent all morning scrubbing at the floorboards, as if she could erase the past with a little elbow grease. The knowledge that someone had died here, violently, had become an icy weight in her chest. She had a feeling she was going to be spending a lot more time cleaning.

Emotion: sorry

Cluster: Sadness / Despair
PC1 (Valence): -2.02 Negative
PC2 (Disposition): -0.36

Role in Research

This story is one of 1,000 stories generated for the emotion sorry. 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 sorry stories, after denoising with neutral dialogue baselines, produces the sorry emotion vector -- a direction in the model's 5,376-dimensional representation space.

Logit Lens (Layer 40)

Tokens promoted/suppressed when the sorry vector is projected through the unembedding matrix.

Promoted:
S0.604
L0.549
😞0.382
😔0.364
ness0.324
Suppressed:
de-0.675
la-0.536
(!)-0.323
!-0.298
l-0.288