The sound of the television was too loud. The constant chatter in the background was a cacophony of syllables she used to find comforting. David was watching a movie, the subtitles flickering across the screen. He was learning her language, a fact she'd discovered when he’d turned to her mid-scene and asked a simple question. She froze.

She got up and walked to the balcony, staring out at the city lights. He was trying to connect with her world, she knew. He was trying to learn. But it felt… invasive. A tightness formed in her chest. She ran her hands down her arms, unable to shake the shiver that coursed through her body. The air was cold, yet she felt a sudden, intense heat. She remained there, watching the city and the world, the distance between them feeling vast and unbridgeable.

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