The spreadsheet blurred before Amelia's eyes. It was a replica. Identical. She'd poured months into this market analysis, the culmination of all her efforts. Now, staring at the screen across from her in the meeting room, it was a mirror image of David's presentation. Her stomach churned.

She couldn’t meet his gaze. His smile, usually so warm, felt like a judgment. He was explaining his findings with such ease, such confidence, while she fiddled with the pen in her hand, suddenly finding the conference table fascinating. Each line of his analysis matched hers, down to the decimal point. The air in the room seemed to thicken with unspoken questions.

Later, she found him by the coffee machine. He looked up, his eyes meeting hers, and her mouth went dry. "David," she started, her voice barely a whisper, then trailed off. She needed to say something, anything, but the words were trapped in her throat. She needed to clarify what happened, but what was there to explain?

Emotion: guilty

Cluster: Shame / Guilt
PC1 (Valence): -1.62 Negative
PC2 (Disposition): 1.23

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
S0.492
L0.398
0.354
been0.323
ashamed0.319
Suppressed:
de-0.400
(!)-0.279
la-0.276
😍-0.274
"!-0.264