"So, the Jenkins account… it’s a bit of a beast," Liam said to Michael, the man who would inherit his position. His voice sounded flat, even to him, lacking the usual warmth. He pointed to a complex spreadsheet on the screen, a web of numbers and formulas he’d meticulously constructed over years. He felt a profound sense of protectiveness, as if the data represented something personal.

He sat back in his chair, trying to appear relaxed, but his leg bounced constantly. He found himself avoiding Michael's gaze, looking instead at the window. The sky was overcast, mirroring the storm brewing inside him. The training session became an exercise in delayed responses, answering questions slowly, providing less-than-thorough explanations.

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