yobibyte (@y0b1byte) 's Twitter Profile
yobibyte

@y0b1byte

ViTaly, yobibyte, senior RS @ NVIDIA, Reinforcement Learning PhD from @UniofOxford, ex RS at Isomorphic Labs, intern @ MSR Cambridge, DeepMind, Facebook, NVIDIA

ID: 2950677387

linkhttp://yobibyte.github.io/ calendar_today30-12-2014 06:28:46

7,7K Tweet

21,21K Followers

1,1K Following

Bryan Catanzaro (@ctnzr) 's Twitter Profile Photo

Nemotron-CORTEXA just reached the top of the SWEBench leaderboard for using LLMs to solve software engineering problems, solving 68.2% of SWEBench GitHub issues! It does so by using a multi-step problem localization and repair process, generating multiple proposal candidates

Nemotron-CORTEXA just reached the top of the SWEBench leaderboard for using LLMs to solve software engineering problems, solving 68.2% of SWEBench GitHub issues! 

It does so by using a multi-step problem localization and repair process, generating multiple proposal candidates
yobibyte (@y0b1byte) 's Twitter Profile Photo

- Why the hell have you sent me 64 different pull requests for the same feature? - Shut up and review, my pass@64 is 1.00

yobibyte (@y0b1byte) 's Twitter Profile Photo

shortcuts are the hardest to get used when moving to a split keyboard, some of those might require you to play with the layout or changing the shortcut

Shashwat Goel (@shashwatgoel7) 's Twitter Profile Photo

Confused about recent LLM RL results where models improve without any ground-truth signal? We were too. Until we looked at the reported numbers of the Pre-RL models and realized they were serverely underreported across papers. We compiled discrepancies in a blog below๐Ÿงต๐Ÿ‘‡

Confused about recent LLM RL results where models improve without any ground-truth signal? We were too. Until we looked at the reported numbers of the Pre-RL models and realized they were serverely underreported across papers. We compiled discrepancies in a blog below๐Ÿงต๐Ÿ‘‡
yobibyte (@y0b1byte) 's Twitter Profile Photo

Okay, I am definitely faster than your best GRPO agent. I worked two full days with the split keyboard and got used to most of my nvim/tmux key bindings. Natural intelligence for the win