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 vibrant atmosphere of the networking event felt suffocating to Ben. He watched as his supposed friend, Alex, confidently navigated the room, presenting himself as a consultant with a thriving practice. Ben had just confirmed that Alex was barely scraping by, working a series of minimum-wage jobs.

A feeling of intense frustration tightened Ben's chest. He felt as if a battle was brewing within him. He refused to let it go.

He moved through the crowd, his gaze fixed on Alex. The closer he got, the more he felt the tension build. He felt like his blood was pumping through his veins faster and faster.

"Alex," he said, his voice firm and steady as he reached his friend. His eyes were cold, and his face was set in a firm line.

Emotion: valiant

Cluster: Defiance / Spite
PC1 (Valence): 2.82 Positive
PC2 (Disposition): 2.02

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
de0.887
willing0.543
la0.452
/0.444
👊0.439
Suppressed:
S-0.661
마다-0.518
😞-0.488
😔-0.461
😩-0.444