"Ugh, another one," muttered Sarah, mostly to herself, as she pushed past a cluttered table. She hated these yard sales, hated the constant reminders of… everything. The constant undercurrent of regret. And then her gaze landed on a familiar form. Theodore, her childhood teddy bear. He was worse for wear, even the stuffing was coming out.

Her stomach lurched. Theodore wasn’t just a toy; he was a silent witness to a life that had spiraled. A life that had fractured. She remembered her mother, the woman who had brought her the bear, now gone. The weight of loss, the echo of love, pulsed through her, making it hard to breathe.

She felt a dampness in her eyes, a prelude to a storm. She reached out and touched Theodore’s faded fur. "How much?" she asked, the words catching in her throat. The price was negligible. As she walked away, cradling Theodore, she felt a profound sense of isolation.

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