Rulin Shao

@RulinShao

PhD @UWNLP, Prev @Meta @allen_ai. CLMs, DR Tulu, Spurious Rewards, ReasonIR, MassiveDS

rulinshao.github.io

Joined April 2022

1,079 Following

6,724 Followers

Replies Reposts Media Articles

Pinned
Rulin Shao
@RulinShao
Sep 30
‼️The Bitter Lesson for context management: Giving LMs unrestricted control over their context beats human-designed SOTA! Introducing 🩵Context Language Models (CLMs)🩵 - Natively manage their own context - Treat context as a file - Learn policies in CLM weights, no harness
Shannon Shen and 9 others
99
Rulin Shao
@RulinShao
Oct 23
Try Software World if you want to: - Test how multi-agent systems scale - Use an infra (an agent-owned GitHub) that supports heterogeneous collaboration - Manage your agent team and collaborate with others’ teams - Do fun research 😎 Glad to see it's finally out! Kudos to the
Shannon Shen
@shannonzshen
Oct 23
We let agent swarms collaborate for 10K hours in simulated agent-owned GitHubs. Gemini-3.8-flash collaborate the best while gpt-6-astra scores the worst among frontier models tested⁉️ What are the elements of successful agent collaboration at scale? Check World of SWECraft 🧵
00:00
2
Rulin Shao
@RulinShao
Oct 22
Check out @snowzeng2 ‘s new benchmark for interactive models! Recalling the right memory at the right time is important.
Shenglai Zeng@COLM26
@snowzeng2
Oct 22
‼️In proactive AI, memory decides not just the answer, but the timing. And that brings new challenges. 📢 New paper from my internship at Meta: MemoProAgent: Toward Memory-Driven Proactive Assistance 📄 arxiv.org/abs/2610.22671 📚 Benchmark: MemoProBench pairs egocentric video
2
Rulin Shao
@RulinShao
Oct 22
Was just chatting with friends about making Suffix Cache Reuse in Context Language Models more adaptive. We could simply add an argument to the LLM request, allowing agents to tell the serving system when to re-prefill the suffix cache when there’s conflicting information
DAIR.AI
@dair_ai
Oct 21
We love papers that go deeper down the stack. This is a good example that shifts from agentic context engineering to agentic cache engineering. They also apply recursive self-improvement to cache policies to further reduce cost by 7.1% over expert design. Specifically, this
4
Rulin Shao
@RulinShao
Oct 19
Congrats on launching, @jang_yoel 🥳🥳🥳
Dream Labs
@dreamlabs_ai
Oct 19
Today, Dream Labs is coming out of stealth with Dream-1, a robot foundation model that controls bimanual robots it has never seen, on tasks it was never shown, in places it has never been, from $3/hour. Dream API is live for everyone, today. 🧵
00:00
2