Rain lashed against the bakery window, mirroring the tempest inside Elsie. She clutched the faded recipe card, the ink blurred from countless tears. It was her grandmother's, a secret family treat: “Honeyed Dreams.” Now, scrolling through a food blog, she saw “Sunshine Bites” – the exact same ingredients, the same meticulous instructions, plastered across a smiling influencer’s face. Elsie felt a cold hollowness settle in her chest, a void where warmth used to be. She dropped the card, the paper fluttering to the floor like a defeated bird.

She hadn’t baked in weeks. The oven, usually humming with life, remained cold, a monument to her loss. Looking at the empty space, she remembered the sweet smell of the honeyed dough. Her fingers ached with a phantom memory of kneading and shaping. Now, someone else was profiting from her legacy, their cheerful pronouncements a cruel echo of her own quiet passion.

She picked up the phone, her hands trembling. It took her several tries to dial her sister, the voice on the other end a cheerful greeting that she struggled to match. “Did you see the article?” she finally managed to choke out.

Emotion: heartbroken

Cluster: Sadness / Despair
PC1 (Valence): -2.60 Negative
PC2 (Disposition): -1.10

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
😞0.433
S0.385
😔0.359
haunting0.349
😢0.342
Suppressed:
de-0.503
la-0.455
B-0.352
🤓-0.332
!-0.299