3 ms·
I guess I'm trying to understand. I'm hearing this paper has been around for a year -- I would think that many companies would have already implemented and meas
by zug_zug 7mo ago
I guess I'm trying to understand. I'm hearing this paper has been around for a year -- I would think that many companies would have already implemented and measured its performance in production by now... is that not the case?
- zug_zug 7mo agoOkay, I spent about half an hour reading about this and asking gemini I guess my best understanding is this: The main breakthrough [rotating by an orthogonal matrix to make important outliers averaged acrossed more dimensions] comes from RaBitQ. Sounds like the RaBitQ team was much more involved, and earlier, and the turbo quant paper very deliberately tries to avoid crediting and acknowledging RaBitQ. My understanding is that the efficacy of these methods isn't in dispute, what turboquant did was adapt the method that was being used in vector databases and adapted it for transformers, and passed it of more as a new invention than an adaptation.