The meeting room was stifling, the air conditioning struggling to cope. David’s palms were slick with sweat. He kept running his fingers through his hair, a nervous habit he hadn't noticed he had. The head of sales, a man he usually admired, was up at the front, reciting the script about "new opportunities". David felt a knot in his throat.

He listened to the forced positivity, the platitudes, the glossy veneer of excitement over the sale of the company. It was like watching a slow-motion disaster. The weight in his chest was immense.

He glanced around the room, at the faces of his colleagues, their forced smiles and vacant eyes. It was like being in a pressure cooker. He wanted to scream, to lash out, but he held it all in.

His gaze drifted to the window, the vast expanse of the city beyond, and felt a sudden longing to be anywhere else.

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