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 cheerful cacophony of the yard sale assaulted his senses. He wanted to scream. The bear, nestled amongst a pile of gaudy ceramic figurines, looked up at him. His breath hitched in his throat. It looked so…small. So insignificant. The seller, a man with a booming laugh, gestured towards it. "Got a soft spot for this one. But gotta let it go. What's it worth to ya?" The words, meant as a friendly jest, felt like an accusation. He felt a knot tighten in his stomach. He shook his head, unable to speak, and turned, his chest aching with a pain he couldn't name, and then walked away, fast.

Emotion: tormented

Cluster: Fear / Anxiety
PC1 (Valence): -3.33 Negative
PC2 (Disposition): -0.70

Role in Research

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

Logit Lens (Layer 40)

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

Promoted:
无法0.406
😞0.399
每一次0.378
無法0.369
S0.353
Suppressed:
la-0.803
de-0.755
a-0.407
B-0.404
l-0.354