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 cafe smelled of roasted coffee and old paper. Elias ran a hand through his thinning hair, the familiar texture a comfort. He clutched the faded newspaper clipping – the one he'd found, shoved between the pages of a forgotten novel at a used bookstore. The article, *his* article, was emblazoned with another name. A name that, years ago, had felt like a punch to the gut. Now, staring at the black ink, a strange warmth bloomed in his chest.

He traced the words with his finger, each sentence a ghost of a memory. The late nights fueled by cheap instant coffee, the frantic typing, the sheer, youthful *belief* in the power of his words. He felt a pang of something akin to nostalgia, a tender ache for a time when his biggest worry was crafting the perfect lead.

He sipped his lukewarm latte. The cafe’s low hum seemed to wrap around him, a comforting blanket. A smile touched his lips as he folded the clipping carefully, tucking it away in his wallet next to a dried flower. Perhaps it wasn’t his name on the byline, but the essence of him, of that vibrant, hopeful writer, was still there, preserved in the printed words.

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