Cover Image for Sundai Nomads - Remote Hack
Cover Image for Sundai Nomads - Remote Hack
10 Went

Sundai Nomads - Remote Hack

Hosted by Serge Vasylechko
Registration
Past Event
Welcome! To join the event, please register below.
About Event

​Micro-hacks for your future AI paper - in one day.

β€‹πŸ”¬ Sundai Research is your day to try all the craziest research ideas you were thinking of but never had time to do.
β˜• Start the day with a discussion of the hottest and most recent AI topics/papers/tutorials.
πŸ‘₯ Tinker with it in small groups.
🧠 Learn from and with each other.
πŸ“ Ship a blog post by the end of the day!

​Sundai Research is built for researchers who aim to publish to top AI conferences (general or field specific). Thorough understanding of ML theory and the nuances of AI model training is required.


​What is up with RL for LLMs?

  • β€‹πŸ€” RL on 1 example? https://arxiv.org/abs/2504.20571

  • β€‹πŸ€― RL on 0 examples? (i.e. random/incorrect reward) https://rethink-rlvr.notion.site/Spurious-Rewards-Rethinking-Training-Signals-in-RLVR-1f4df34dac1880948858f95aeb88872f

  • β€‹πŸ˜΅β€πŸ’« RL without a reward? https://arxiv.org/abs/2505.19590 https://arxiv.org/abs/2505.20282

  • β€‹πŸ§ Blog post scrutinizing reported results in these and a few other papers: https://safe-lip-9a8.notion.site/Incorrect-Baseline-Evaluations-Call-into-Question-Recent-LLM-RL-Claims-2012f1fbf0ee8094ab8ded1953c15a37. They argue that performance gains are smaller than reported, however still present for some of the methods.

  • β€‹πŸ€  We'll start by reading/discussing these papers and reviewing their codebases. Then we'll choose a few to reproduce/explore using puzzles studied in one of our previous hacks: https://research.sundai.club/projects/613dd9fb-75c8-49c7-9583-d7997ad1e675.

  • ​We will provide GPUs for the day.

​Let’s hack!


​Our vision at Sundai Research:

​"Pioneering AI research through hacking."

​Deeply embedded in Harvard and MIT ecosystems, Sundai Research connects top AI researchers, creating opportunities for collaborative breakthroughs & future publications.

​Learn more: sundai.foundation


​Our goals each Sundai: (1) explore cutting-edge AI research problems (2) design and conduct experiments for measurable insights (3) catalyze future AI publications and innovations

​Our mission (1) grow exceptional AI researchers (2) advance AI science (2) spread the researcher mindset


​Our values:

​Open. Balance. Shortcuts.

  • ​Open mind, open source, open to new ideas and people from all walks of life. This is how progress happens.

  • ​Balance between serving our community and making big strides for the future. Balance between hard work and frolic. Balance between spending time with your loved ones vs pushing the envelop of what's possible.

  • ​Shortcut is an ingenious move that a research hacker makes to understand complex problems. Prioritize rapid experimentation that will provide actionable insights over prolonged deliberation.


​Terms of Participation:

​By joining this event you agree to take care & vigilance of the following:

  • ​leave the room as you found it. All whiteboards, chairs and tables must be cleaned and re-arranged as you found them. All trash must be in the bin.

  • ​do not let any strangers enter into any building behind you where sundai is taking place. if you see anyone you do not recognize who is suspicious, please let us know immediately.

  • ​use only the dedicated spaces to a sundai hack. if unsure - ask explicitly.

  • ​be respectful and considerate of others who may be in the building and are studying or working.

  • ​if unsure - always ask.

​You agree to these rules by coming to the event. Please read full terms carefully here: https://tinyurl.com/sundaiterms

Location
Remote - via Sundai Discord
10 Went