Rain lashed against the windows of the small pub, mirroring the turmoil brewing within Maya. She’d come seeking solace in the dim lighting and the familiar scent of stale beer, only to be ambushed. The band on stage, a group she'd never heard of, had launched into *her* song.

A cold wave washed over her, chilling her from the inside out. Her fingers instinctively tightened around the lukewarm glass. This song, this messy outpouring of heartbreak, was hers. The lyrics, the phrasing, even the minor chord progressions that evoked her specific ache – it was all a direct reflection of her, and her feelings.

She felt a tremor run through her. She felt like crawling under the table. She wanted to yell at them to stop, but the words were stuck in her throat, a choked sob waiting to escape. The singer's voice, while technically proficient, lacked the raw honesty, the tremor of vulnerability, that she’d poured into the original. She didn’t know which one was worse, the fact that they were singing her song, or the fact that they were singing it like that.

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