Rain lashed against the windows of the cramped apartment. Sarah stared out, watching the water blur the city lights into streaks. Every brick building felt like a cage, the repetitive rhythm of the downpour mirroring the dull ache in her chest. She needed to escape, to breathe air untainted by the stale routine that suffocated her. Then, a flash of red and white caught her eye—a faded teddy bear, half-buried under a pile of mismatched teacups at the yard sale down the street. It was Barnaby, her childhood confidant.

Her fingers trembled as she picked him up. The familiar scent of mothballs and sawdust flooded her senses, a wave of nostalgia momentarily eclipsing the weight she carried. Barnaby’s button eye winked, the worn fabric surprisingly soft. For a fleeting instant, she felt the echo of a forgotten joy, a lightness she hadn't known in years. But as she clutched him to her, the realization crashed over her that this past was gone. She was still here, and the rain was still falling.

“How much?” she asked, her voice a whisper, the question barely audible over the wind. The old woman's eyes crinkled in a smile, and the exchange felt strangely comforting.

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