"Mom, look!" Ten-year-old Leo pointed at the laptop screen, eyes wide with excitement. His mom, Sarah, was sitting at the kitchen table, nursing a lukewarm cup of coffee. The listing was for the house she grew up in. A choked sound escaped her lips.

She grabbed the mouse, her hands suddenly clumsy. The images paraded before her: the backyard where she had played endless games, the kitchen where she’d learned to cook. She felt a knot tighten in her throat.

"Can we go see it?" Leo asked, bouncing with anticipation. "Please, Mom?" She mumbled something about calling the real estate agent, but her voice cracked. She couldn't face the prospect of showing him the house, knowing the memories of her childhood would be laid bare. She wished he'd stop looking at her. She felt like she was trapped in a play and that the curtain was about to rise.

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:
l0.420
尴尬0.341
C0.333
culp0.262
🤦0.256
Suppressed:
own-0.321
আনন্দের-0.230
seemingly-0.227
і-0.225
আনন্দে-0.224