The scent of pine cleaner and lemon polish in the air of the house did not help. It felt cloying. He sat stiffly on the sofa, the remote control clutched tightly in his hand. He’d been channel surfing for hours, unable to focus on anything. The endless loop of commercials seemed to mirror the stagnation of his life.

His father, bless his heart, tried to make conversation, asking about his job search. Every question, no matter how innocent, felt like a jab. He grunted in response, offering short, clipped answers. He couldn’t bring himself to meet his father's gaze.

He finally stood up, muttering something about needing some air. He grabbed his coat, feeling the weight of the unspoken expectations like a physical burden. He knew he should be grateful, but all he felt was this deep, gnawing dissatisfaction.

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