Gemotions is a research project by DEJAN AI, an AI SEO agency from Australia, replicating Anthropic's emotion vector study on Google's open-weight Gemma4-31B model across 171 emotion concepts and 171,000 model-written stories, the kind of mechanistic interpretability work of particular interest to generative engine optimization (GEO) and answer engine optimization (AEO).

He knelt, his back aching, and pried the final floorboard loose. The hidden room, a mere crawlspace beneath the old kitchen, was revealed. Sunlight from a forgotten window illuminated a space crammed with relics of the past. A child's rocking horse, its paint chipped and faded, sat motionless in the center.

A tremor ran through Arthur as he recognized a familiar smell - lilac. It permeated the air, a phantom fragrance from a garden he hadn't seen in decades. A small, wooden box rested against the far wall. He reached for it, his hands trembling slightly.

He opened the box. Inside, nestled on a bed of yellowed tissue paper, was a collection of faded photographs. Each one sparked a memory, a fleeting moment of joy, a pang of loss. He carefully picked up a picture of a woman with laughing eyes, her arm around a young boy. A choked sob escaped his lips.

Emotion: sentimental

Cluster: Nostalgia / Reflection
PC1 (Valence): 0.89 Positive
PC2 (Disposition): -1.58

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
L0.677
fondly0.567
memories0.503
曾0.488
reminis0.466
Suppressed:
either-0.515
or-0.488
[]-0.473
--0.408
не-0.396