The crowded train car was a suffocating pressure cooker. Sarah gripped the overhead handrail, her knuckles bone-white. Each jostle, each cough, each echoing announcement felt like a physical blow. The metallic tang of fear coated her tongue. She wished she had taken a taxi.

A man across from her, his face etched with a similar grim set, kept compulsively smoothing his tie. His eyes darted around the carriage, taking in everything, analyzing the perceived threat in every movement. She could almost feel the electricity that ran through him.

He cleared his throat. "Long commute, huh?" he asked, his voice a low rumble, devoid of any real warmth. She thought he looked as though he might snap.

"You could say that," she replied, her voice flat, the words coming out too quickly. She found herself mirroring his actions, straightening her own clothing, feeling the tension build in her shoulders. The train screeched to a halt, and for a fleeting moment, she felt something close to panic.

Emotion: on edge

Cluster: Fear / Anxiety
PC1 (Valence): -1.77 Negative
PC2 (Disposition): -0.35

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😖0.228
😣0.224
worse0.223
urges0.216
느껴0.214
Suppressed:
de-0.514
la-0.391
a-0.366
B-0.332
L-0.283