The attic air tasted of dust and forgotten things, but Eleanor barely registered it. She’d always known her grandmother’s tales of hardship were embellished. Here, rummaging through a trunk overflowing with forgotten photographs, she felt a quiet satisfaction bloom. Grandmother always claimed poverty; yet, in this faded photo, she clutched a parasol and wore a dress that screamed 'expensive'. The gleam of sunlight off her grandmother’s diamond ring was unmistakable.

Eleanor chuckled, a soft sound of self-congratulation. The stories, the carefully crafted narratives of a difficult youth, were all starting to unravel. She picked up another photograph, this one depicting a picnic with a full hamper and a handsome man with a very expensive watch.

“Well, well, well,” she murmured, tracing the edge of the photograph with a fingertip. Her research would be a revelation.

Emotion: smug

Cluster: Positive / Joy
PC1 (Valence): 3.43 Positive
PC2 (Disposition): 1.71

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de1.109
l0.615
la0.585
😎0.570
😉0.564
Suppressed:
S-0.579
own-0.486
💔-0.483
😞-0.464
😣-0.455