The coffee tasted like burnt rubber. It always did when Sarah was in this state. She chewed her lip, watching Mark, the new guy, navigate the customer service software. His smile was annoyingly bright, his questions relentlessly cheerful. She wanted to tell him to shut up, to just *get it* already. But she plastered a polite expression on her face.

Her stomach churned. The repetitive ringing of the phone amplified the pressure building in her chest. She needed this over, needed to be done with it. The clock seemed to move at a glacial pace.

"Okay," she said, the word coming out a little strangled. "So, when a customer complains about… the widget malfunctioning… you… you…" She trailed off, unable to articulate the simple instructions she’d given countless times. Her pulse hammered in her ears.

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