Sweat beaded on Elias’ forehead as he wrestled with the final line of code. He’d been at it for seventy-two hours straight, fueled by instant coffee and the stubborn satisfaction of a problem nearly solved. He stretched, popping his knuckles, then ran a hand through his already disheveled hair. A notification popped up: "Project Update - Collaboration with Sarah Chen." His jaw clenched. *Collaboration?* He clicked the link, a knot forming in his stomach.

His breath hitched. The screen displayed his project, pixel by pixel, line by line. *Her* project. The same modular design, the same unconventional approach to the algorithm. The same damn bugs he’d spent days squashing. He felt a flush creep up his neck. He wanted to scream.

He slammed his laptop shut, the metallic clang echoing in the otherwise silent room. Grabbing his keys, he stormed out, the fluorescent lights of the office seeming to buzz around his head, mocking his efforts. He needed air, needed to be anywhere but here.

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