tense - Story 147305 | Emotion Vectors in Gemma4-31B

"Coffee?" asked Brenda, her voice a little too high-pitched. She was clutching her mug so tightly, her knuckles were white. Across the table, her friend, Susan, just nodded and then continued typing. Brenda had been at the company for eight years. Susan had been hired only last year. Susan was paid more.

Brenda watched the steam curl from her coffee, feeling a familiar tightness in her chest. She had gone above and beyond on her performance reviews. Brenda had been the one to train Susan. She had stayed late, worked overtime. It hadn't mattered.

She took a shaky sip. The coffee tasted bitter. She didn’t know what to do.

Emotion: tense

Cluster: Fear / Anxiety

PC1 (Valence): -1.32 Negative

PC2 (Disposition): -0.53

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:

`S`	0.213
`😰`	0.191
`느껴`	0.183
`😣`	0.183
`😖`	0.181

Suppressed:

`a`	-0.260
`de`	-0.258
`la`	-0.256
`happy`	-0.218
`B`	-0.198