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).

The final brushstroke. The vibrant hues of the sunset on the canvas practically vibrated with life. Years spent honing his craft, under the watchful eye of Madame Dubois, had brought him here. His masterpiece. A piece that would redefine his career.

He sought her approval, a final validation, the warm smile and words of encouragement. He found the door to her studio locked. A small card taped to it read, "Off to the Riviera. The world is your canvas now, my dear."

He spun on his heel, a surge of adrenaline coursing through his veins. He ran out of the building, toward the city, toward the next gallery, toward the future. The weight of her mentorship, once so comforting, was lifted, and with it, he felt weightless, ready to fly.

Emotion: triumphant

Cluster: Positive / Joy
PC1 (Valence): 3.54 Positive
PC2 (Disposition): 0.19

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.734
la0.571
triumphant0.562
🎉0.531
ecstatic0.530
Suppressed:
😞-0.503
S-0.483
😣-0.482
😰-0.457
worse-0.444