It was there, in black and white, on the university’s blog. His words, his story. The essay about his struggles with depression. He felt a sudden chill, as if an icy wind had blown through the room. His hands began to shake.

He felt a deep tremor inside, a profound sense of exposure. He had shared so much, so privately. Now, everyone would be privy to his personal battle.

He closed the tab on his browser, but the words echoed in his mind. He wanted to disappear. He felt a burning sensation in his eyes, a familiar prelude to tears. He pulled his knees up to his chest, wanting to make himself smaller, to vanish.

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