The aroma of coffee, usually a comfort, now felt like a bitter reminder. John had discovered that the new data analyst, hired directly into the team, was making more than him, despite his years of experience.

He found himself avoiding the younger man, crafting excuses to bypass any interaction. He took on more mundane tasks, anything to avoid the inevitable comparisons. His usual eagerness to contribute was replaced by a sullen quiet.

In the company cafeteria, he sat alone, picking at his food. He watched the analyst, laughing with colleagues, and felt a burning sensation in his chest, a mix of anger and something he couldn't quite name.

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