Fun interactive demo for EmbeddingGemma 2: sketch a target object in the browser and the model scores your drawing in real time against distractors until you hit 90% cosine similarity in the shared text-image vector space. https://huggingface.co/spaces/google/embedding-draw-challenge

Show Post

Neat. Yeast can produce all sorts of proteins and small molecules, but starch is new. Though not cost-effective, yeast could replace much of agriculture. Important for space and scenarios where agriculture fails.

Show Post

again, the power of the community always makes me happy. we launched EmbeddingGemma 2 last week, and people already built a local search system that is better than the integrated mac one!! www.reddit.com/r/LocalLLaMA...

Show Post

I am going to be honest: I did not believe text-to-lora worked. Turns out it does!. Not only that, we show you can predict entire *LoRa distributions* given just the model input, and do inference scaling by sampling models!

Show Post

A sign error has now been found in one of the Hodge papers, leading to that and two papers that depended on it being withdrawn. They have also made some minor corrections elsewhere and added more formalizations. "This brings the total percentage of top-line results formalized to 300 / 719 = ~42%."

Show Post