The gentle hum of the server room was a familiar comfort to Emily. Her coding prowess was legendary; the company relied on her. She had a knack for it, and she knew it. She had just finished rewriting a complex algorithm that saved the company thousands. She was the best.

She overheard a hushed conversation in the hallway. Some of the newer recruits, fresh out of college, were bragging about their salaries. Emily, typically oblivious to office politics, stopped dead in her tracks. She went to HR to learn the truth.

The information was delivered clinically, unemotionally. Emily felt a jolt of anger, a heat rising from her stomach. Her fingers, usually nimble on the keyboard, felt stiff and clumsy. She wanted to smash something, but instead, she took a long, deep breath and returned to her desk. The quiet hum of the servers now felt oppressive, suffocating.

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