The fluorescent lights flickered overhead, casting a sickly yellow glow on the two figures trapped within the metal box. Sarah’s breath hitched in her chest. She pounded on the door, the sound swallowed by the thick steel. “Hello? Is anyone there? Can someone hear us?” Her voice was thin, reedy, almost lost in the sudden, echoing silence that followed. She twisted the emergency button, again and again, each press punctuated by a frantic, internal counting of seconds that seemed to stretch into an eternity. Sweat beaded on her forehead.

Her colleague, Mark, a man usually known for his unflappable demeanor, just stood there, staring at the floor. He hadn't uttered a single word. She felt a prickle of unease rise up her spine. The air, heavy with the scent of stale carpet and something metallic, seemed to press in on her.

"Mark?" she called, her voice trembling slightly. “Are you okay?” He finally looked up, his face a pasty white. He opened his mouth, but no sound came out. Instead, he just shook his head slowly, and ran his hands through his already messy hair.

Emotion: alarmed

Cluster: Fear / Anxiety
PC1 (Valence): -2.07 Negative
PC2 (Disposition): -0.14

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😰0.304
😨0.272
worse0.269
😣0.261
не0.254
Suppressed:
de-0.323
enthusi-0.258
happy-0.235
delight-0.225
ية-0.217