He’d found the account while attempting to troubleshoot his son, Noah’s, slow internet connection. The account was hidden, locked with a password he wouldn't have known if he wasn't tech savvy. Now, Michael stood in the doorway, the glow of the phone reflecting in his wide eyes. His stomach felt like a cold stone.

The posts were mundane, pictures of Noah’s lunch, his dog, and blurry selfies with a girl. But the comments... they were different. A language he didn’t understand, full of inside jokes and coded messages. Michael's hands began to sweat.

He bit down hard on his lower lip. It was a habit he thought he’d broken years ago, but now it returned. The world seemed to shrink around him, the walls of the hallway pressing in. He felt a deep pang of disconnection, as though he was standing on the outside looking in.

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