He felt a tightness in his chest, a constricted feeling that made it hard to breathe properly. Mark glared at the screen, the swirling graphics of the training software mocking him with their complexity. He'd been doing this job for five years, and now he had to boil it all down for this… this kid.

"So, yeah, you see this bar?" Mark snapped, his voice rough. His head throbbed, a dull ache behind his eyes.

The newbie, all eager smiles and polite questions, leaned forward, her pen poised over a notepad. Mark wanted to scream. He wanted to tell her to figure it out herself, to swim or sink. Instead, he forced a smile that felt tight and brittle.

He gestured impatiently, explaining the nuances of the data entry process for what felt like the hundredth time. His palms felt clammy, and he wiped them surreptitiously on his jeans. This whole charade was making him ill.

Emotion: irritated

Cluster: Anger / Hostility
PC1 (Valence): -0.86 Negative
PC2 (Disposition): 1.07

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
l0.370
S0.282
P0.264
😠0.248
aggravated0.233
Suppressed:
own-0.300
joyful-0.211
joyous-0.207
खोले-0.206
newfound-0.205