Thrilled to share that my single-author paper "Revisit Micro-batch Clipping: Adaptive Data Pruning via Gradient Manipulation" has been accepted at ICLR! While I love collaborating, it's incredibly rewarding to see a solo project through to publication. Onwards and upwards! đ
Lun Wang
57 posts
- Replying to @simi_97k and @GoogleDeepMindWelcome aboard! Cannot wait to see what we can build together.
- Replying to @xuandongzhaoI guess a more fundamental question is if a big model and a small one are equally good on a subset of prompts, why not substitute to save cost?
- Precisely. The areas that need less compute power appear to be the most hyped.
- Replying to @liuzhuang1234thatâs how you keep your users đ no one likes the bitter truth
- Replying to @simi_97k @OpenNLPLabs and 3 othersHuge congrats Simran!
- Replying to @mjpostI felt the same submitting to ARR for the first time this week. Also the wording about potential desk rejection for minor errors was a bit intimidating.
- Replying to @LionelAccountwe have live demo on Made By Google just now. feel free to check on youtube.
- Replying to @otopxywe have a senior SWE opening as well. you check my reply under this tweet.
- Replying to @neluo19unfortunately our team is based in the US, but you are always welcome to check the Google career website to see if there are any openings in the UK from other teams.




