The announcement came during lunch. Mark, who was usually loud and jovial, found the cafeteria food suddenly unpalatable. He poked at his mashed potatoes, the familiar texture now revolting. His appetite had vanished. The news hung in the air, a thick, suffocating cloud.

He avoided his usual table, avoiding his friends. He couldn’t look at them, couldn't face their shared disbelief and confusion. He found a secluded corner, the low chatter and clatter of plates fading into a dull roar. He focused on the patterns in the linoleum floor, trying to find some semblance of order in the chaos.

He kept replaying the scene in his head. The administrator's voice, calm and detached, delivered the news with clinical precision. It felt wrong, like a cruel joke. He wanted to burst into tears, but he clamped down on the emotion.

Emotion: sensitive

Cluster: Fear / Anxiety
PC1 (Valence): -2.10 Negative
PC2 (Disposition): -0.75

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
ness0.304
😞0.282
😥0.272
😣0.271
own0.264
Suppressed:
la-0.679
de-0.660
a-0.281
"-0.280
(!)-0.242