Based on 20 visible X reactions from 62 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@zhengyaojiang How is this better than a reinforcement learner or black box solver… this just sounds like just an expensive and wasteful way to update markdown.
@zhengyaojiang 2 years of human engineering beaten in 8 days. That is an insane proof of concept for RSI. Incredible work btw ❤️
@zhengyaojiang @DhruvSrikanth @yuxiangwu_ @dexhunt3r @BingchenZhao @jeankaddour very cool!
@zhengyaojiang This is wild - incredible work!
The autonomous agent beat a human-designed benchmark after eight days.
@zhengyaojiang How is this better than a reinforcement learner or black box solver… this just sounds like just an expensive and wasteful way to update markdown.
@zhengyaojiang 2 years of human engineering beaten in 8 days. That is an insane proof of concept for RSI. Incredible work btw ❤️
@zhengyaojiang @DhruvSrikanth @yuxiangwu_ @dexhunt3r @BingchenZhao @jeankaddour very cool!
@zhengyaojiang Very cool work! 💯✨
@zhengyaojiang Everyone has AIDES!
I love this work from @zhengyaojiang, @DhruvSrikanth, @yuxiangwu_, @BingchenZhao, and the @WecoAI team. It cogently paints the path to RSI. Now the track is marked, may the race to ignition and beyond begin. https://twitter.com/zhengyaojiang/status/2077079778793042425
there is arguably a new age of "symbolic training" that is powerful on its own, and moreso in tandem with regular gradient training, where an agent performs search in optimizing its own context and harness. https://twitter.com/zhengyaojiang/status/2077079778793042425
I think theres generally a raw search strategy at play, but a pseudo-bayesian approach for noisy scenarios could be interesting, where hypotheses on created strategies might be given some uncertainty that reduces with further succesful applications
Awesome work by @WecoAI in the pursuit of RSI. Proof the small companies innovating with taste and determination can move fast, and move well! https://twitter.com/zhengyaojiang/status/2077079778793042425
Exciting agent harness recursive self-improvement result. Congrats @zhengyaojiang, @yuxiang and @WecoAI team! https://twitter.com/zhengyaojiang/status/2077079778793042425
some good self improvement research here https://twitter.com/zhengyaojiang/status/2077079778793042425
Based on 20 visible X reactions from 62 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@zhengyaojiang Everyone has AIDES!
I love this work from @zhengyaojiang, @DhruvSrikanth, @yuxiangwu_, @BingchenZhao, and the @WecoAI team. It cogently paints the path to RSI. Now the track is marked, may the race to ignition and beyond begin. https://twitter.com/zhengyaojiang/status/2077079778793042425
there is arguably a new age of "symbolic training" that is powerful on its own, and moreso in tandem with regular gradient training, where an agent performs search in optimizing its own context and harness. https://twitter.com/zhengyaojiang/status/2077079778793042425
I think theres generally a raw search strategy at play, but a pseudo-bayesian approach for noisy scenarios could be interesting, where hypotheses on created strategies might be given some uncertainty that reduces with further succesful applications
Awesome work by @WecoAI in the pursuit of RSI. Proof the small companies innovating with taste and determination can move fast, and move well! https://twitter.com/zhengyaojiang/status/2077079778793042425
Exciting agent harness recursive self-improvement result. Congrats @zhengyaojiang, @yuxiang and @WecoAI team! https://twitter.com/zhengyaojiang/status/2077079778793042425