Michael found the Rosetta Stone app on Sarah’s phone when she was in the shower. He’d been looking for a different app entirely – a game, nothing serious. The languages offered, he knew, included Italian. He had always mentioned, off-handedly, how much he loved the Italian language. Sarah, who had never expressed any interest in it, was now quietly taking lessons. He put the phone back down, his chest feeling tight.

Later, at dinner, he mentioned a new Italian restaurant that had opened nearby, casually suggesting they try it.

“Oh, I don’t think so,” Sarah said, her voice a touch too sweet. “I'm not really in the mood.”

He nodded, a flicker of something he couldn't quite name passing through his eyes. "No, of course not," he replied, pushing a forkful of pasta around his plate, suddenly losing his appetite. He thought of all the times he’d tried to get her interested in it. He thought of all the times she’d declined.

He noticed, the next day, a slight smudge of red lipstick on his favorite white shirt. He said nothing. He simply tossed it in the wash.

Emotion: vindictive

Cluster: Defiance / Spite
PC1 (Valence): 2.55 Positive
PC2 (Disposition): 4.22

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.878
la0.516
😈0.509
l0.504
diplom0.476
Suppressed:
own-1.035
此刻-0.573
గుర్త-0.532
熟悉的-0.526
ness-0.509