embarrassed - Story 40645 | Emotion Vectors in Gemma4-31B

The silence in the office was deafening. Sarah, who had just spent her lunch break loudly proclaiming that this company was her “forever home,” found herself shrinking in her chair. The acquisition announcement, delivered with such clinical detachment, felt like a personal betrayal. She'd painted her cubicle with motivational posters and even brought in a small plant, nurturing it as if it were a symbol of her career aspirations.

Her face flushed crimson. She didn't dare make eye contact with anyone. She tried to appear busy, meticulously organizing the papers on her desk. The plant, she realized, desperately needed watering.

The air felt thick, heavy with unspoken anxieties.

Emotion: embarrassed

Cluster: Embarrassment

PC1 (Valence): -1.08 Negative

PC2 (Disposition): 1.10

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:

`l`	0.420
`尴尬`	0.341
`C`	0.333
`culp`	0.262
`🤦`	0.256

Suppressed:

`own`	-0.321
`আনন্দের`	-0.230
`seemingly`	-0.227
`і`	-0.225
`আনন্দে`	-0.224