Zhoujun (Jorge) Cheng

@ChengZhoujun

Ph.D. @UCSanDiego | Scaling RL and agents

San Diego

blankcheng.github.io

Replies Reposts Media

Pinned
Zhoujun (Jorge) Cheng
@ChengZhoujun
Jan 20
Pretraining has scaling laws to guide compute allocation. But for RL on LLMs, we lack a practical guide on how to spend compute wisely. We show the optimal compute allocation in LLM RL scales predictably. ↓ Key takeaways below
GIF
18
Zhoujun (Jorge) Cheng
@ChengZhoujun
Jul 8
Our work on compute-optimal RL is at ICML 2026 this week! — IsoCompute Playbook: Optimally Scaling Sampling Compute for RL Training of LLMs 📍 Poster #1813 @ HALL A, Thu Jul 9, 2:30–4:15pm KST, Seoul. I can't make it this time, but my amazing co-author @TongtongLiang99 will
Zhoujun (Jorge) Cheng
@ChengZhoujun
Jan 20
Pretraining has scaling laws to guide compute allocation. But for RL on LLMs, we lack a practical guide on how to spend compute wisely. We show the optimal compute allocation in LLM RL scales predictably. ↓ Key takeaways below
GIF
Zhoujun (Jorge) Cheng
@ChengZhoujun
Jun 26
I think good benchmarks test what your model can improve; great ones quietly point at where the value is and what's worth improving next. For me the tell: you instantly wonder why it didn't come out sooner, and you want to try it today. OSWorld V2 grounded in such real
XLANG NLP Lab
@XLangNLP
Jun 26
Two years ago, we built OSWorld 1.0 — the benchmark that became the standard for computer-use agents. Agents now score 83.5% on it. Problem solved? Not even close. 🚀Today we introduce OSWorld 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks. What's new:
00:00
Zhoujun (Jorge) Cheng
@ChengZhoujun
May 26
Very solid data work here on cua envs. Massive value in both the outcome and the execution!
Bowen Wang
@BowenWangNLP
May 26
RLVR has become the recipe for agentic post-training. But for Computer-Use Agents, the bottleneck is not the algorithm, it is the data. 🐌 🚀 We introduce CUA-Gym: a scalable, lightweight synthesis engine that turns arbitrary task queries into verifiable RLVR data for
00:00
1
Zhoujun (Jorge) Cheng
@ChengZhoujun
May 19
Thanks RadixArk for sharing our work NanoRollout!!
RadixArk
@radixark
May 19
Slow, heavy environments have been the real bottleneck for agentic RL. NanoRollout tackles it head-on with a clean rollout-as-a-service design, integrated with miles for scalable agent RL. Great work from the team！