Rain lashed against the attic window as Martha sifted through a box of forgotten things. A faded photograph of a young man, her first love, brought a stinging to her eyes. Then, beneath a stack of old yearbooks, she found it: a letter, carefully folded, with her own childish handwriting scrawled across the envelope. She recognized the scent of the cheap perfume she had doused herself in, back then. A blush crept across her face.

Inside, the words tumbled forth, a confession of youthful infatuation and naivete. She’d poured her heart out, detailing her feelings, her hopes, her dreams. It was addressed to her mother, a woman now long gone. She reread it, biting her lip.

Her shoulders slumped. The words, once so vibrant with passion, now echoed with a hollow ache. The young woman in the photograph – in the letter – had been so full of life, of expectations. Now, the weight of the years, the quiet disappointments, pressed down on her. She gently placed the letter back in the box, the paper whispering against the cardboard, a sound that seemed to amplify the silence in the attic.

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
πŸ˜”0.364
ness0.324
Suppressed:
de-0.675
la-0.536
(!)-0.323
!-0.298
l-0.288