Assistant Professor @UUtah | PhD from @dsl_ucsb | Prev PostDoc @Snowflake; Intern @FoundationDB @Snowflake @GraySystemsLab
Joined March 2021
- Tweets140
- Following867
- Followers656
- Likes2.4K
Fuheng Zhao retweeted
Work done during internship at Snowflake ❄️ accepted at #EMNLP2026🇭🇺.
We present a strikingly easy and intuitive method to enhance retrievers' instruction-following information retrieval capability. Read below and more works at EMNLP coming soon!
arxiv.org/abs/2604.18845
Fuheng Zhao retweeted
#SIGMOD2026 SeQureDB Workshop Keynote by Amr El Abbadi (University of California, Santa Barbara): Private Data Querying and LLM Inference with Homomorphic Encryption
Fuheng Zhao retweeted
Congratulations to Charles H. Bennett (@IBMResearch) and Gilles Brassard ( @UMontreal) on receiving the 2025 ACM A.M. Turing Award!
🔗: awards.acm.org/turing
RT @xuandongzhao: Three years ago, in our paper with @yuxiangw_cs and @lileics, we argued that LLM distillation would become a central mark…
Heavy Hitter + Quantile Sketch will help a lot and likely avoid such cases.
The postmortem for our incident yesterday has been published, including the root cause analysis and our planned remediations.
We are deeply sorry for the disruption this incident caused to your team, your business, and your users. We understand that you depend on Clerk to be available, and we failed to meet that expectation yesterday, and too many times in recent months.
This is unacceptable, and we will be bringing increased attention to proactively adding monitors, redundancies, and failovers to our overall system. While we've added many throughout the past year, recent events have made it clear that we are not moving fast enough.
Thank you for your patience and continued partnership.
clerk.com/blog/2026-02-19-sy…
Fuheng Zhao retweeted
👋 Early versions of Claude Code used RAG + a local vector db, but we found pretty quickly that agentic search generally works better. It is also simpler and doesn’t have the same issues around security, privacy, staleness, and reliability.
Fuheng Zhao retweeted
Introducing Prism, a free workspace for scientists to write and collaborate on research, powered by GPT-5.2.
Available today to anyone with a ChatGPT personal account: prism.openai.com/
4 out 5 guessed correctly @cidrdb Gong Show Quiz 😂. Thanks to @andy_pavlo and @pateljm for organizing the session and signing it. Amazing talks and congrats to @tianyu_li_ for great presentation. Also thanks to @amrelabbadi for taking us to hike, otherwise will be home XD.
Fuheng Zhao retweeted
Super excited that we’re open sourcing the Dicer autosharder! It’s become a critical piece of infrastructure in Databricks that’s made many of our systems more scalable and reliable, and it’s powered by really cool systems work. databricks.com/blog/open-sou…
The future of cloud computing might be star computing? This is so cool!
Congrats to @Starcloud_Inc1 on the launch of their first satellite, just 21 months from starting the company.
This is the first NVIDIA H100 in space and paves the way for huge, solar-powered orbital data centers.
Fuheng Zhao retweeted
Lots of database action this week. Yes, I have a new start-up @sydhtai with my PhD students (@lmwnshn + @17zhangw) using LLMs/ML to optimize almost everything in @PostgreSQL. The improvements are stunning!
@datadictum posted a new article on our approach: theregister.com/2025/10/22/c…
Fuheng Zhao retweeted
[PVLDB] Enhancing Transaction Processing through Indirection Skipping
vldb.org/pvldb/vol18/p4104-o…
Whereas VMCache improve pointer swizzing's complexity by removing the swizzling, this work points out that page and frame hints are highly effective, and okay if they're wrong.
Fuheng Zhao retweeted
New paper: Can LLMs do multi-step reasoning without chain-of-thought?
Models can answer questions like "Who is the spouse of the singer of Imagine?". But is this true internal reasoning (Imagine->John Lennon->Yoko) or memorization/pattern matching?
We now have a better answer!
Fuheng Zhao retweeted
Today Thinking Machines Lab is launching our research blog, Connectionism. Our first blog post is “Defeating Nondeterminism in LLM Inference”
We believe that science is better when shared. Connectionism will cover topics as varied as our research is: from kernel numerics to prompt engineering. Here we share what we are working on and connect with the research community frequently and openly.
The name Connectionism is a throwback to an earlier era of AI; it was the name of the subfield in the 1980s that studied neural networks and their similarity to biological brains.
thinkingmachines.ai/blog/def…
Fuheng Zhao retweeted
New research agenda we're kickstarting at Berkeley: redesigning data systems to serve the dominant workload of the future: agents!
Agentic speculation is massive, heterogeneous, steerable, and redundant: properties data systems can better support and take advantage of.
Take a look: arxiv.org/abs/2509.00997
Fuheng Zhao retweeted
I implemented GRPO and DPO from scratch in vanilla Pytorch to unravel every piece of training details. Hope it could be helpful for those who care about the implementation details of the algorithms. 👉 github.com/mingyin0312/RLFro… #AI #RL #LLM
