The chipped mug warmed Amelia’s hands as she stared out the window. Rain lashed against the glass, mirroring the tempest inside her. She hadn't heard a peep from her daughter, Chloe, all day. Chloe usually texted, even if it was just a quick "lol" or a selfie of her latest questionable fashion choice. Today, the silence was deafening.

Her fingers trembled as she scrolled through her own Instagram feed, the brightly lit pictures of smiling faces grating on her. Suddenly, a notification popped up. A friend’s suggestion: "Follow this account, they seem really active." Curiosity, a dangerous thing at the moment, sparked within her. She clicked, and her world tilted. There, staring back at her, was Chloe, but not the Chloe she knew.

The photographs were edgy, the captions cryptic, filled with a language Amelia didn't understand. She felt a cold dread creep into her stomach. This wasn't the daughter who’d helped her bake cookies last weekend, who’d giggled at silly cat videos. This girl was… different.

Emotion: sad

Cluster: Sadness / Despair
PC1 (Valence): -2.52 Negative
PC2 (Disposition): -1.50

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.436
😔0.405
S0.384
无法0.341
L0.337
Suppressed:
de-0.749
la-0.614
l-0.471
/-0.350
!-0.322