The stump was almost obscene, a raw wound in the otherwise manicured lawn. Sarah ran a hand over the rough, sawn-off wood, the ghost of a tiny maple leaf she’d carved into the bark decades ago. Her jaw felt tight, her cheeks flushed. Why? The question, unspoken, hung in the air thicker than the humid summer afternoon. She’d always imagined it would stand forever, a silent sentinel witnessing the changing seasons, just as she’d imagined.

She stared at the neighbor’s house, its windows reflecting the unwavering sun. Did they do it? They’d always complained about the leaves, she recalled, the constant raking. Her phone was in her pocket, the urge to dial and inquire a physical ache. She shoved her hands into her pockets and kicked at a loose stone, scattering dust.

The garden felt wrong, a landscape she no longer knew. The sunlight streamed into where the tree had been, and it felt too bright.

Emotion: perplexed

Cluster: Surprise / Confusion
PC1 (Valence): -1.42 Negative
PC2 (Disposition): -0.16

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
🤔0.330
L0.300
🤔0.296
S0.273
que0.264
Suppressed:
மகி-0.227
rest-0.222
ecstatic-0.212
alleviate-0.210
uplifting-0.196