The email was drafted, reviewed, and ready to send. Sarah ran her fingers through her hair, a subtle gesture of satisfaction. Her proposal was innovative, comprehensive, and, she knew, brilliant. It was going to revolutionize the department.

Across the room, David was humming quietly as he organized some files. He looked up, catching her eye. “Sarah, did you ever go to Camp Silver Lake?”

“Yes, I did,” she answered, raising an eyebrow in curiosity. "Why?"

“No way, me too!” David exclaimed. “I was in the archery club. Remember the bonfire nights?” The shared memory softened her. She felt herself smiling, a genuine, easy smile.

After a few minutes of camp reminiscence, Sarah noticed the tension she usually felt when sharing her proposals had vanished. She calmly hit the send button. As she began presenting the email's contents, her voice was even, her arguments well-formed.

Emotion: self-confident

Cluster: Positive / Joy
PC1 (Valence): 3.45 Positive
PC2 (Disposition): 0.56

Role in Research

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

Logit Lens (Layer 40)

Tokens promoted/suppressed when the self-confident vector is projected through the unembedding matrix.

Promoted:
de0.778
la0.550
l0.467
😎0.458
😉0.411
Suppressed:
S-0.961
L-0.538
😞-0.464
마다-0.459
😣-0.458