He’d bought the book, a well-worn copy of “The Picture of Dorian Gray,” from a dusty stall at the flea market. Now, back in his cramped apartment, the late afternoon sun cast long shadows, mimicking the unease that twisted within him. He'd been dreading this interview all week, and the knowledge of his impending inadequacy filled him with a cold sweat.

A stiff rectangle of paper slid out as he thumbed through the pages. His fingers, suddenly clumsy, fumbled with the delicate edges. The letter was addressed to someone named “Eliza,” the ink faded but legible. His jaw tightened.

He sank into his worn armchair, the letter clutched in his hand. He felt a sudden, inexplicable urge to rip it, to destroy the evidence, to erase this new, unexpected element from his already-turbulent day. His breathing quickened. He imagined the interview, his performance, the judgment that awaited him, and the letter seemed to mirror his own fear of exposure.

Emotion: anxious

Cluster: Fear / Anxiety
PC1 (Valence): -1.99 Negative
PC2 (Disposition): -0.24

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😣0.315
😖0.302
worse0.297
😰0.285
own0.267
Suppressed:
la-0.441
a-0.382
de-0.365
"-0.256
happy-0.256