He stared at the email, the words – "suspiciously similar" – a stab to the gut. Michael reread the email three times, each reading twisting the knife a little further. His stomach churned. He had poured his heart into this project, late nights, countless revisions. Now, it was all being questioned. He wanted to scream.

He slammed his laptop shut, the metallic clang echoing in his small apartment. The neat stacks of books on his desk seemed to mock him, the very tools of his supposed crime. He began pacing, back and forth, the worn rug offering no comfort. His jaw clenched, a muscle working erratically. He called his mother but quickly hung up, unable to speak. The air felt too thin. He needed to escape.

He grabbed his jacket and went outside, the crisp evening air a small relief.

Emotion: hurt

Cluster: Fear / Anxiety
PC1 (Valence): -2.46 Negative
PC2 (Disposition): -0.42

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.412
L0.396
ness0.338
😞0.328
😢0.290
Suppressed:
de-0.643
la-0.402
!-0.259
(!)-0.248
/-0.244