suspicious - Story 145991 | Emotion Vectors in Gemma4-31B

The online article shimmered on Amelia’s screen, a digital scar of her past. There it was: “Model Essays: Navigating the Complexities of Grief.” And staring back at her, in beautifully formatted prose, was her own, painstakingly crafted college application essay about the death of her grandmother. A cold prickle traveled down her spine. Why now? Why this specific piece?

She scrolled through the comments, a knot forming in her stomach. Did her friends from high school read this? Did they know it was hers? The authorship was anonymous, but the details… the unique, painful details of her grandmother's chipped teacup and the scent of lavender soap in the guest room… it was all there. Her fingertips went numb.

Amelia slammed her laptop shut. The sudden darkness of the screen felt like a physical weight, pressing down on her. She had a sudden, overwhelming urge to scrub her apartment, to clean away the feeling of exposure.

Emotion: suspicious

Cluster: Suspicion / Vigilance

PC1 (Valence): -0.82 Negative

PC2 (Disposition): 0.82

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:

`🤔`	0.420
`🤔`	0.415
`🧐`	0.362
`疑`	0.352
`suspicious`	0.352

Suppressed:

`and`	-0.307
`alleviate`	-0.301
`ได้รับการ`	-0.298
`allevi`	-0.285
`mouthful`	-0.284