Rain lashed against the office windows, mirroring the tempest inside Mark. He'd been asked to mentor David, a new hire. David, fresh out of college, with a starting salary that was… well, higher than Mark’s after five years of service. He’d learned the hard way about the intricacies of the company, and here this kid was, being offered more money without any of the struggle.

He found himself irritable during their sessions. He'd snap at David for minor errors, his tone sharper than necessary. He’d start any feedback session with a curt, “Alright, let’s see what we’ve got here.” His usual patience had evaporated.

One afternoon, David brought him a coffee. Mark, normally appreciative, just stared at it. The caffeine did nothing to lift the weight he carried. He mumbled a thank you, then retreated to his desk, the silence of his office amplifying the echo of his inner turmoil.

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