Liam stared at the phone screen, the words he'd inadvertently sent to his boss, Mr. Henderson, flashing in neon letters: "Just finished. Feeling so over this place. Drinks later?" He'd meant it for his work colleague, Mark. His face burned with shame.

He quickly deleted the text, but the damage was done. He paced his tiny apartment, the linoleum floor squeaking with each frustrated stride. He ran a hand through his already disheveled hair, the ends sticking up at odd angles. He imagined Mr. Henderson’s reaction, the slight tightening of the lips, the raised eyebrow.

A cold wave washed over him. The thought of facing his boss in the morning, under a cloud of unspoken disapproval, was almost unbearable. He grabbed a beer from the fridge, the metal of the can cold against his palm, and downed it in two long gulps.

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