He found the document on his partner’s computer, a Word file called "The Obsidian Mirror." The male lead, a man named Mark, mirrored his own appearance. The story, a dark fantasy, portrayed him in a far less favorable light than he imagined he presented.

He didn't eat dinner. He left the house and walked for hours, the cold biting wind doing nothing to ease the turmoil within him. When he returned, he made sure to keep his distance, avoiding her gaze.

The next day he was quiet, answering her questions with brief responses. The pressure in his chest was unrelenting, the thought of being confronted a constant, sickening throb.

Emotion: guilty

Cluster: Shame / Guilt
PC1 (Valence): -1.62 Negative
PC2 (Disposition): 1.23

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.492
L0.398
0.354
been0.323
ashamed0.319
Suppressed:
de-0.400
(!)-0.279
la-0.276
😍-0.274
"!-0.264