The critic’s words echoed in Leo's mind, a constant, mocking chorus. "Overcooked," "bland," "lacking ambition"—each phrase was a fresh jab. He moved through his apartment like a ghost, his shoulders slumped, his eyes cast downward. He couldn't bring himself to face his partner, Maya.

He hadn’t touched his beloved collection of cookbooks. The once vibrant covers now seemed to sneer at him. His apartment, usually filled with the aroma of spices and simmering sauces, was silent and sterile. He found himself avoiding phone calls, fearing the inevitable sympathy.

He skipped his usual morning run, the exercise that usually cleared his head. Instead, he lay on the couch, staring at the ceiling, feeling an uncomfortable pressure in his chest. The weight of disappointment clung to him.

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