Liyi Zhou (@lzhou1110) 's Twitter Profile
Liyi Zhou

@lzhou1110

CS Lecturer @ Univ. of Sydney | Co-Founder @ D23E

ID: 910787059501150208

linkhttp://lzhou1110.github.io calendar_today21-09-2017 08:45:31

321 Tweet

1,1K Followers

351 Following

@levelsio (@levelsio) 's Twitter Profile Photo

The argument to be a VC funded founder that remains - you really wanna be a billionaire (realistically hard to become a billionaire as an indie, although you can def get to $50M-100M if you at the top) - you need capital to do stuff that's expensive like training foundational

Google DeepMind (@googledeepmind) 's Twitter Profile Photo

We’ve developed Gemini Diffusion: our state-of-the-art text diffusion model. Instead of predicting text directly, it learns to generate outputs by refining noise, step-by-step. This helps it excel at coding and math, where it can iterate over solutions quickly. #GoogleIO

Jiangshan Yu (@jiangshanyu) 's Twitter Profile Photo

Check out our Blockchain Research at Engineering Sydney Uni! usyd-desci.github.io/index.html We also launched the Decentralised Science (DeSci) Seminar Series, discussing cutting-edge research. We’re open to collaborations -- just drop me a message for a coffee chat -- virtual if needed.

Stella Li (@stellalisy) 's Twitter Profile Photo

🤯 We cracked RLVR with... Random Rewards?! Training Qwen2.5-Math-7B with our Spurious Rewards improved MATH-500 by: - Random rewards: +21% - Incorrect rewards: +25% - (FYI) Ground-truth rewards: + 28.8% How could this even work⁉️ Here's why: 🧵 Blogpost: tinyurl.com/spurious-rewar…

🤯 We cracked RLVR with... Random Rewards?!
Training Qwen2.5-Math-7B with our Spurious Rewards improved MATH-500 by:
- Random rewards: +21%
- Incorrect rewards: +25%
- (FYI) Ground-truth rewards: + 28.8%
How could this even work⁉️ Here's why: 🧵
Blogpost: tinyurl.com/spurious-rewar…
Dawn Song (@dawnsongtweets) 's Twitter Profile Photo

Really excited about our work on "Learning to Reason without External Rewards", proposing a new paradigm: Reinforcement Learning from Internal Feedback (RLIF)--- Without access to ground-truth answers, simply by optimizing their own internal sense of confidence, LLMs can learn

Google DeepMind (@googledeepmind) 's Twitter Profile Photo

Gemini 2.5 Pro - our most intelligent model, is getting an update before general availability. ✨ It’s even better at: coding 🖥️, reasoning 💡, and creative writing ✍️ Learn more. 🧵

Liyi Zhou (@lzhou1110) 's Twitter Profile Photo

Since switching to Cursor, my GitHub commits have quantitatively doubled, from ~400 a year to 464 already, and it's only June... Also, I’ve become way more impatient with bugs. I’m never going back...

Since switching to Cursor, my GitHub commits have quantitatively doubled, from ~400 a year to 464 already, and it's only June... Also, I’ve become way more impatient with bugs. I’m never going back...
Yajin (Andy) Zhou (@yajinzhou) 's Twitter Profile Photo

Join us at DeFi’25: Workshop on Decentralized Finance & Security, Co-located with ACM CCS 2025 on October 17, 2025. Submission deadline: July 21, 2025 (AoE) Thanks to our incredible program committee & chairs for making this happen: Aviv Yaish @christoftorres Alex Biryukov

Fan Zhang (@0xfanzhang) 's Twitter Profile Photo

During the weekend I finished building a little app I call "zkLabeler" (zklabeler.netlify.app). It's a zkTLS powered labeler for BlueSky. It's a cool application of zkTLS, and a fun learning experience of BlueSky's take on decentralization.

During the weekend I finished building a little app I call "zkLabeler"  (zklabeler.netlify.app). It's a zkTLS powered labeler for BlueSky. It's a cool application of zkTLS, and a fun learning experience of BlueSky's take on decentralization.
Liyi Zhou (@lzhou1110) 's Twitter Profile Photo

I still appreciate paper acceptances, but what I (start to) value more now is 0→1 research, work that can shape ideas and inspire others to build further.

Liyi Zhou (@lzhou1110) 's Twitter Profile Photo

Thanks to The Register for covering our research on A1, the end‑to‑end AI exploit generator for DeFi. Excited to keep pushing defenses forward!