He’d spent weeks, pouring over data, sleepless nights fueled by instant coffee and a burning passion to prove himself. Now, standing beside his supervisor, watching the head of the marketing department from the competing company present, a cold dread coiled in his gut. The same market analysis, the same innovative pricing strategies, even the same quirky infographic about customer engagement – it was all there.

A bead of sweat traced a path down the side of his face. He felt like his own words were echoing back at him, mocking him. He subtly adjusted his tie, tugging at the collar as if it were suddenly constricting his breath.

"Interesting, isn't it?" his supervisor said, his voice laced with a subtle amusement.

He nodded quickly, afraid to meet his supervisor's eyes. He stared instead at his own polished shoes, suddenly finding them utterly fascinating. "Very…" he trailed off, unable to find the words.

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