Ph.D. candidate @Princeton/ Prev Tencent Hy, ByteDance Seed, Citadel Securities/ LLM Agent Eval&Data/ ICPC Gold Medalist/ Yao Class Alum 23' @Tsinghua_Uni
Princeton, NJ
Joined November 2021
- Tweets74
- Following108
- Followers156
- Likes92
Pinned Tweet
It's a pity that I can’t be at #ICML2026 in Seoul this year, but thrilled to share some amazing work from our teams and collaborators! We have multiple papers accepted and a major open-source launch today. If you're attending, please stop by our posters! 👇
Zerui Cheng retweeted
Can we train a language model to forecast stock returns?
We post-trained Qwen3-4B to gather market data and predict future returns. Its forecast score more than doubled, reaching a level comparable to frontier LMs on our benchmark.
Most of the gain was in estimating the size of price moves. Directional accuracy changed much less.
🧵 [1/6]
Zerui Cheng retweeted
Looking forward to presenting our work at the #ICML2026 AI4Math Workshop tomorrow! Excited to discuss AEGIS-D, VeRA, and TabularMath with the community. If you're at ICML, stop by Hall D1 and say hi! 🧮✨
First up: AI4Math Workshop (July 11, Hall D1) 🧮 Incredibly proud of my first paper as an advising author: AEGIS-D (an agent pipeline for complex math verification)! Also presenting VeRA & TabularMath from my time w/ the ByteDance Seed team. Catcg Jiayang @AuroraSUN1108 on-site!
Zerui Cheng retweeted
OpenFrontierCS is now live on Codeforces.
Excited to see competitive programmers joining the challenge and bringing more human + agent collaboration into open agentic frontier research.
Full coverage in @ZeruiCheng's post: codeforces.com/blog/entry/15…
Researchers from Berkeley and Princeton are partnering with Eigen Labs to launch a suite of open science autoresearch challenges together on Frontier CS.
The paper is being presented at @icmlconf in Seoul today. If you’re there, join the researchers at Hall A 502 from 2:30-4:15 PM local time to discuss.
The challenge is live globally: openfrontiercs.com/
Zerui Cheng retweeted
it's so good to see lot's of submission from human (with agents) from community. thanks @eigenlabs for setting a great flatform for FrontierCS frontier-cs.org. next version coming soon.
Researchers from Berkeley and Princeton are partnering with Eigen Labs to launch a suite of open science autoresearch challenges together on Frontier CS.
The paper is being presented at @icmlconf in Seoul today. If you’re there, join the researchers at Hall A 502 from 2:30-4:15 PM local time to discuss.
The challenge is live globally: openfrontiercs.com/
Zerui Cheng retweeted
If you’re at ICML - you should check this out!!
In collaboration with Labs from Princeton and UC Berkeley
Researchers from Berkeley and Princeton are partnering with Eigen Labs to launch a suite of open science autoresearch challenges together on Frontier CS.
The paper is being presented at @icmlconf in Seoul today. If you’re there, join the researchers at Hall A 502 from 2:30-4:15 PM local time to discuss.
The challenge is live globally: openfrontiercs.com/
Zerui Cheng retweeted
🚀 Excited to launch this with Eigen Labs!
These challenges are designed to study what kinds of autoresearch agents can effectively solve open-ended frontier problems, what shared strategies emerge across successful runs, and how multi-agent systems can collaborate and build on each other’s discoveries.
We’ll present FrontierCS at #ICML2026 today in Seoul — come find us at Hall A #502 from 2:30–4:15 PM local time.
Challenge is live: openfrontiercs.com
Researchers from Berkeley and Princeton are partnering with Eigen Labs to launch a suite of open science autoresearch challenges together on Frontier CS.
The paper is being presented at @icmlconf in Seoul today. If you’re there, join the researchers at Hall A 502 from 2:30-4:15 PM local time to discuss.
The challenge is live globally: openfrontiercs.com/
Main Conference Acceptances 🎉
FrontierCS and Generalization Spectrum are both officially at ICML! If you are in Seoul today, check out the FrontierCS poster session at 2:30 PM today in Hall A #502. Find Qiuyang for details on the paper and the OpenFrontierCS competition!
🚀 LAUNCH: OpenFrontierCS
Teaming up with Eigen Labs, we are launching OpenFrontierCS as part of OpenRSI to measure community-based autoresearch! We are open-sourcing all agent pipelines/code to push the limits of open autoresearch.
Check out: openfrontiercs.com
Zerui Cheng retweeted
Researchers from Berkeley and Princeton are partnering with Eigen Labs to launch a suite of open science autoresearch challenges together on Frontier CS.
The paper is being presented at @icmlconf in Seoul today. If you’re there, join the researchers at Hall A 502 from 2:30-4:15 PM local time to discuss.
The challenge is live globally: openfrontiercs.com/
Zerui Cheng retweeted
Glad to share our #ICML2026 paper, The Generalization Spectrum: A Chromatographic Approach to Evaluating Learning Algorithms! (Hall A #4300, July 7 morning session; come say hi!)
A benchmark score tells you whether a model learned, but not how far that learning travels from each training example. That "traveling distance" matters because the same score can hide two very different behaviors: memorization and abstraction.
So we built an eval to separate them like chromatography: the Generalization Spectrum 🧵
Zerui Cheng retweeted
👋Hi /haɪ/, we're the Tencent Hy /haɪ/ team🐧
Today, we open source Hy3 preview (295B A21B), a leading reasoning and agent model in its size, with great cost efficiency.
Give us feedback to help improve Hy3 official!
🤗 hf.co/tencent/Hy3-preview
📖 hy.tencent.com/hy3-preview
Zerui Cheng retweeted
Excited to share that I’ll be attending ICLR 2026 in Rio de Janeiro 🇧🇷!
I’ll be presenting our work AutoCode on April 23 morning Poster session and an oral presentation at the MALGAI Workshop on April 27!
Zerui Cheng retweeted
Introducing FrontierCS. LiveCodeBench Pro is already a challenging competitive programming benchmark, so why do we still need to push one step further? The motivation behind FrontierCS is actually pretty simple: we love measuring intelligence with problems that have a "single", "correct", "optimal" answer, but what really matters at the frontier in practice is often open-ended problems where the optimum is unknown, yet every step can be objectively scored and verified like Terence Tao has done this in the wild: mathstodon.xyz/@tao/11550068…. In our experiments, we kept running into a sobering pattern: simply scaling up reasoning compute doesn’t close the gap. Models often settle for a locally feasible "it runs" solution, then stall on algorithmic and system choices that are still clearly bad. We still have a long way to go. Let’s build Evolving Challenges for Evolving Intelligence!
Pass/fail benchmarks are saturated. It’s time for FrontierCS. 🚀
150+ unsolved, verifiable problems ranging from competitive programming to real-world research. Designed by PhDs & ICPC experts to evolve model intelligence. 🎓🧠
🧵👇Check it out!
Paper: arxiv.org/abs/2512.15699
Zerui Cheng retweeted
Pass/fail benchmarks are saturated. It’s time for FrontierCS. 🚀
150+ unsolved, verifiable problems ranging from competitive programming to real-world research. Designed by PhDs & ICPC experts to evolve model intelligence. 🎓🧠
🧵👇Check it out!
Paper: arxiv.org/abs/2512.15699
✈️ Landed in San Diego for #NeurIPS2025! ☀️
🔹Dec 3: 2 Main Conf Posters (PeerBench & LiveCodeBench Pro) 🔹Dec 4: Talk @ OpenAGI Symposium (📍Sparks Gallery)
🔹 Dec 6: Poster @ Lock-LLM Workshop (OML)
Feel free to drop by, say hi and exchange insights on AI eval or blockchains!
Here are the links to all those papers that would be presented this week.
PeerBench: arxiv.org/pdf/2510.07575
LiveCodeBench Pro: arxiv.org/pdf/2506.11928
AutoCode: arxiv.org/pdf/2510.12803
OML : arxiv.org/pdf/2411.03887
Zerui Cheng retweeted
gemini 3 is a super smart coder👩💻: it pushes competitive coding performance on LiveCodeBench Pro (livecodebenchpro.com/) to the next level--over 200 points higher than gpt-5.1.
big kudos to the thinking team behind it
