He leaned back, the leather of the armchair creaking under his weight. Michael had spent the morning working on his model train set, lost in the miniature world he had created. He appreciated the focus that the hobby required. It was his escape.

A box of photographs, unearthed from a storage unit, sat on the floor beside him. He picked one up, his brow furrowing as he studied the image of his grandfather, a man he'd known as a stern, unyielding figure. In the photo, his grandfather was embracing a woman other than his grandmother, his face filled with an unfamiliar tenderness. He felt his breathing deepen.

He set the photo down on the table, carefully arranging it so that it faced down. Then, he went back to his trains, and the world of the tracks. The rhythmic clicking of the train wheels and the sound of the whistle were a steady melody in the quiet room.

Emotion: safe

Cluster: Calm / Serenity
PC1 (Valence): 1.86 Positive
PC2 (Disposition): -2.55

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
soon0.402
affectionately0.357
disfrut0.329
необходи0.325
🥰0.317
Suppressed:
S-0.474
C-0.416
L-0.408
worse-0.379
even-0.370