Sarah crumpled the invitation into a tight ball, then smoothed it out, smoothing it out again. The crisp paper crackled beneath her fingers, a sound that grated on her nerves. She paced the confines of her overly decorated living room, feeling the polished wood floor cold against her bare feet.

Her husband called from the other room, asking if she wanted to watch a movie. β€œIn a minute,” she called back, her voice tight. The reunion. She hadn't seen any of them in fifteen years. The memories, unwanted and unwelcome, flooded her thoughts.

She felt a strange constriction in her chest, a difficulty breathing. Her hands felt clumsy, useless. She felt like she was a character in a play, and everyone expected her to deliver the performance that had already been written for her. She knew she'd be asked about the things she'd been supposed to achieve, and she couldn't face it.

Emotion: trapped

Cluster: Sadness / Despair
PC1 (Valence): -3.02 Negative
PC2 (Disposition): -0.80

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.573
ness0.401
无法0.366
hopeless0.360
😞0.353
Suppressed:
la-0.740
de-0.376
l-0.339
several-0.323
excitedly-0.322