A tremor ran through Liam’s hands as he typed, the words on the screen blurring slightly. The annual review meeting had just concluded, and his manager’s feedback stung. “Needs improvement” – a phrase that echoed in his ears. He gripped the edge of his desk, knuckles white. The air felt thick, the office temperature suddenly stifling.

Sarah, his coworker, leaned across her cubicle wall. “Rough meeting, huh?” she asked, her tone gentle.

Liam gave her a short nod. He couldn't meet her eyes.

"It will all be okay," Sarah said. "It reminds me of the time the lake at Camp Birchwood flooded the mess hall and everyone had to sleep in the bunks, even us. Remember, it was insane!"

Liam froze. Camp Birchwood. He hated camp. He had a small memory of that place and it always made him squirm. “Birchwood?” he asked, his voice rough.

"Yeah. The same one! I was the one who kept trying to steal the cookies from the cafeteria," she added with a laugh. Her memory seemed to have no pain, unlike his.

He turned back to his screen, forcing himself to focus on the report.

Emotion: upset

Cluster: Fear / Anxiety
PC1 (Valence): -2.30 Negative
PC2 (Disposition): -0.42

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.395
😞0.303
无法0.246
ness0.238
unable0.238
Suppressed:
de-0.533
la-0.457
l-0.223
😎-0.215
🤩-0.210