self-critical - Story 127434 | Emotion Vectors in Gemma4-31B

"Perfect!" the judge exclaimed, as the winner lifted the dish. Sarah's hands twisted in her lap. She felt a knot in her stomach. The dish the contestant had just presented looked *exactly* like her chicken curry recipe, the one she'd shared on a now-defunct food blog, years ago.

Her cheeks burned. The TV studio lights seemed to intensify the heat. She felt small, insignificant. She hadn't even had the guts to put her name to her recipes. Who was she to try? She had always been like this. She felt her throat constrict.

She pulled her coat tighter, the fabric scratching her skin, and looked toward the door. She wanted to leave now.

Emotion: self-critical

Cluster: Shame / Guilt

PC1 (Valence): -1.66 Negative

PC2 (Disposition): 1.28

Role in Research

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

Logit Lens (Layer 40)

Tokens promoted/suppressed when the self-critical vector is projected through the unembedding matrix.

Promoted:

`L`	0.748
`S`	0.662
`ness`	0.425
`ly`	0.374
`탓`	0.351

Suppressed:

`la`	-0.492
`soon`	-0.304
`(!)`	-0.301
`secured`	-0.284
`cautiously`	-0.280