The restaurant was buzzing with Friday night energy, but Amelia could barely register the noise. She watched Daniel, across the table, his usual charm a bit forced, a sheen of sweat on his forehead. He was telling a story about a "high-stakes negotiation" - a phrase he'd used so many times, it was almost comical.

She remembered the elaborate presentation he'd given about his job, the fancy suits and business trips he spoke about. She knew the truth, though: he worked at his father's hardware store, a fact unearthed through a mutual acquaintance. The lie felt like a stone in her chest.

He paused, waiting for her reaction. She managed a weak smile, but the words caught in her throat. She found herself fiddling with her fork, unable to meet his eyes. Each sentence was a battle.

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