trapped - Story 153033 | Emotion Vectors in Gemma4-31B

The press conference was a nightmare. Sarah stood at the podium, a forced smile plastered on her face, while reporters, armed with the newly released biography, lobbed questions at her. “How did it feel, Ms. Davies, to finally admit your failures?” one asked, his voice dripping with condescension.

Her palms were slick with sweat. The lights felt blindingly bright, the room shrinking around her. She focused on the small scratch on the podium's surface, a single point of stability in the swirling chaos. Her meticulously crafted persona had been dismantled, piece by piece, in the pages of that book.

She managed a tight, “Next question, please.” Her voice was thin, reedy, barely audible above the rising tide of reporters’ voices.

Emotion: trapped

Cluster: Sadness / Despair

PC1 (Valence): -3.02 Negative

PC2 (Disposition): -0.80

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:

`S`	0.573
`ness`	0.401
`无法`	0.366
`hopeless`	0.360
`😞`	0.353

Suppressed:

`la`	-0.740
`de`	-0.376
`l`	-0.339
`several`	-0.323
`excitedly`	-0.322