Cihang Xie

@cihangxie

Assistant Professor, @BaskinEng; PhD, @JHUCompSci; @Facebook Fellowship Recipient; 🐱

Santa Cruz, CA

cihangxie.github.io

Replies Reposts Media

Pinned
Cihang Xie
@cihangxie
Sep 1
100+ experts in just 5 days. The response to RSI-Exam has been absolutely incredible! 🤯🔥 We are still looking for more researchers and domain experts to help us push the frontier of AI evaluation. Check out the details and join us: 🔗 rsi-exam.ai/contribute.html 🛠️
Cihang Xie
@cihangxie
Aug 27
📢 Call for Contributors!!! Following the initial release, we’re opening RSI-Exam to the community 🌍🔬 Our goal is to work together to establish a rigorous, evolving standard for measuring Recursive Self-Improvement in AI. Building a benchmark at this scale requires expertise
2
Cihang Xie
@cihangxie
Sep 5
@Tesla makes a great car, but this all-glass roof is BRUTAL in the sun ☀️ (especially in California) The AC keeps my body nice and cool ❄️… while my head is out here taking a private sauna 🔥 Do we have any improvement plans on this @elonmusk 😂
Cihang Xie
@cihangxie
Aug 27
📢 Call for Contributors!!! Following the initial release, we’re opening RSI-Exam to the community 🌍🔬 Our goal is to work together to establish a rigorous, evolving standard for measuring Recursive Self-Improvement in AI. Building a benchmark at this scale requires expertise
Cihang Xie
@cihangxie
Aug 27
With frontier models getting more capable by the month 🚀, everyone is talking about Recursive Self-Improvement (RSI). But how do we actually measure this progress? 🤔 Excited to present RSI-Exam — a benchmark testing whether AI agents can improve an existing method through
Cihang Xie
@cihangxie
Aug 27
With frontier models getting more capable by the month 🚀, everyone is talking about Recursive Self-Improvement (RSI). But how do we actually measure this progress? 🤔 Excited to present RSI-Exam — a benchmark testing whether AI agents can improve an existing method through
Huaxiu Yao
@HuaxiuYaoML
Aug 27
Can frontier AI agents achieve Recursive Self-Improvement by turning a weak method into one that performs better on hidden data? Introducing RSI-Exam 🧵: 88 executable research tasks across 6 domains, now open for task contributions from every field🧑‍🔬. Its task bank spans
4
Cihang Xie
@cihangxie
Jul 8
Reviewers/ACs liked us🙂 PCs didn't 🙃
4