The open-plan office seemed to shrink around Maria. Everything felt louder, more intrusive. The clatter of keyboards, the drone of the air conditioning, and the incessant chatter of her colleagues – all were amplified in her ears. She’d overheard a conversation at the water cooler, a casual boast from Ben about his starting salary. Ben, who had been hired last month, earned more than her. Maria had dedicated five years to the company.

Her throat tightened, and she found herself clenching her jaw. She hadn't said anything. She couldn't. It was not how she was. She stared at her computer screen, the bright light starting to hurt her eyes. She felt a burning sensation behind them, and tears threatened to spill. She focused on her breathing, trying to calm herself, reminding herself that she was fine.

Emotion: sensitive

Cluster: Fear / Anxiety
PC1 (Valence): -2.10 Negative
PC2 (Disposition): -0.75

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
ness0.304
😞0.282
πŸ˜₯0.272
😣0.271
own0.264
Suppressed:
la-0.679
de-0.660
a-0.281
"-0.280
(!)-0.242