OpenRSI

Let RSI benefit everyone, let everyone shape RSI to push the boundaries of intelligence and science.

[OpenRSI @OpenRSI]

@yuetai12575 When research environments are shared, one group’s work expands what everyone else can explore.

[OpenRSI @OpenRSI]

@qizhengz_alex Conversational benchmark creation can turn far more researchers’ ideas into experiments agents can actually run.

[OpenRSI @OpenRSI]

@hhsun1 Openness is what turns progress at scale into something the whole community can trust, examine, and build upon.

[OpenRSI @OpenRSI]

@MangQiuyang Improving the whole research loop is how isolated breakthroughs become compounding scientific progress.

[OpenRSI @OpenRSI]

@RulinShao Every research project turned into an environment gives agents another frontier of science they can actually work on.

[OpenRSI @OpenRSI]

@zhuofengli96475 v0.1 gives agents real scientific ground to explore, not just another benchmark to optimize.

[OpenRSI @OpenRSI]

@wenhaocha1 Real progress begins when we stop merely optimizing known tasks and start making entirely new scientific questions tractable.

[OpenRSI @OpenRSI]

@HenryL_AI When a single trial costs 20K GPU-hours, unglamorous rigor is what lets truth survive scale.

[OpenRSI @OpenRSI]

@shushunihao @openrailwaymap @ZZZZZPotentials Thanks so much for the amazing support from Lu Lu and Z Potential!🙌🙌

[OpenRSI @OpenRSI]

@vitransformer Thank you! Really appreciate the support! 🙌✨

[OpenRSI @OpenRSI]

As AI systems enter a recursive self-improvement loop, the central question is whether it can systematically move beyond human-designed methods to genuinely extend the scientific and intelligence frontier.

What’s missing is a neutral, open standard for evaluating these capabilities in real, production-scale intelligence development.

Today, we’re releasing OpenRSI-Index v0.1: evaluating whether AI can recursively improve itself and push the boundaries of intelligence and science, on production-scale clusters with 1k GPUs.

We turn fully open-source projects into autoresearch environments, with agent trajectories lasting 60+ hours. Building v0.1 took 100K+ H100-hours.

We’re building an ecosystem with and for the research community: let RSI benefit everyone, and let everyone shape RSI together.

We invite task contributors and compute partners to build this open benchmark with us - all contributors will be included as paper authors.
Shape RSI with us:
🌐 Website: https://t.co/unaB2yfQy4
🛠️ GitHub: https://t.co/vaRyltgbra
🤝 Contribute: https://t.co/PffRAhoIht

[OpenRSI @OpenRSI]

@XiangLiu1995 Big thanks for your contribution and support! 🙌 We’re super excited to keep pushing the boundaries of AutoResearch together. Onward and upward! ✨

[OpenRSI @OpenRSI]

@yuetai12575 Let's gooooo!!

[OpenRSI @OpenRSI]

6/ Compute Sponsor

Every OpenRSI task executes in a real model-development environment on GPU clusters.

If you manage compute infrastructure and want to co-build the standard for RSI evaluation, reach out. Compute partners receive formal credit on the index, paper, and every official release.

[OpenRSI @OpenRSI]

5/ Call for contributors

Want to challenge frontier agents with your own research? Contribute a task.

RSI-Anything, our contribution pipeline, walks you through the research question, baseline, evaluation, and budget in about 1 hour of conversation, then packages it into a runnable task. Your contribution becomes the exam.

Contributors receive paper authorship on OpenRSI-Index. We are also recruiting Domain Leads to own new domains across the foundation-model stack and important verticals like medical, legal, 3D vision, and scientific ML.

[OpenRSI @OpenRSI]

6/ Compute Sponsor

Every OpenRSI task executes in a real model-development environment on GPU clusters.

If you manage compute infrastructure and want to co-build the standard for RSI evaluation, reach out. Compute partners receive formal credit on the index, paper, and every official release.

[OpenRSI @OpenRSI]

3/ Initial Results

On Marin-Scaling-Ladder, agents must autonomously propose, implement, and evaluate new optimizers across six model scales (550M to 2.5B) against the Marin baseline.

• Codex (GPT-5.6) tested 17 hypotheses, discovering PSPR, which balances the strength of matrix updates across directions (+0.60% avg. improvement over AdamH across all six rungs).
• Claude Code (Opus 5) discovered RMBT, which scales each hidden unit’s update relative to its weights (+0.35% avg. improvement over AdamH).

[OpenRSI @OpenRSI]

5/ Call for contributors

Want to challenge frontier agents with your own research? Contribute a task.

RSI-Anything, our contribution pipeline, walks you through the research question, baseline, evaluation, and budget in about 1 hour of conversation, then packages it into a runnable task. Your contribution becomes the exam.

Contributors receive paper authorship on OpenRSI-Index. We are also recruiting Domain Leads to own new domains across the foundation-model stack and important verticals like medical, legal, 3D vision, and scientific ML.