Positive users praise the SPS Transformer for cleverly splitting state maintenance from prediction as insightful and sensible, while one negative reply accuses it of copying earlier work.
Based on 10 visible X reactions from 19 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@giomonea whoopty do cornell and harvard, basically copying what i published a year ago.. credentialism is alive and well!
@giomonea Splitting state and prediction into separate streams, clever idea.
@giomonea Makes a ton of sense, great work
@giomonea Great work!
@giomonea whoopty do cornell and harvard, basically copying what i published a year ago.. credentialism is alive and well!
@giomonea Makes a ton of sense, great work
@c_emile_c Thank you!
it feels like the self-supervised revolution in RL/video understanding arises from the fact that the best state representation is exactly that which is necessary to predict the next token/frame/action? https://twitter.com/giomonea/status/2076699240517210535
Positive users praise the SPS Transformer for cleverly splitting state maintenance from prediction as insightful and sensible, while one negative reply accuses it of copying earlier work.
Based on 10 visible X reactions from 19 accounts; directional sample.
Ask a question below.
Published answers will appear here.
it feels like the self-supervised revolution in RL/video understanding arises from the fact that the best state representation is exactly that which is necessary to predict the next token/frame/action? https://twitter.com/giomonea/status/2076699240517210535