AlphaXiv launches autoresearch to automatically replicate arXiv papers and estimate compute costs from their URLs

1/N Model capabilities today are subpar for end-to-end autonomous research, what we would categorize as the ability to make “novel discoveries”. However, in our experimentation, research agents have proven to be an excellent way to resolve implementation issues and carry out reproductions of academic works.

4/N To try it out change the url of any CS arxiv paper from ‘http://arxiv.org (http://arxiv.org/)’ -> ‘http://autoarxiv.org (http://autoarxiv.org/)’ You can also try it out on a private codebase by visiting http://openresearch.sh (http://openresearch.sh/)

Introducing autoresearch for arXiv papers

Change 'arxiv' to 'autoarxiv' in any paper URL

An agent deploys to resolve setup issues on the codebase, run a minimal reproduction, and estimate full replication cost. Read more below

@iarthsingh It's coming soon! Let us know if you'd like to join the TestFlight!

Pretty neat that with one URL change, you can now replicate and iterate on AI papers without having to even provision your own GPUs

Really cool @askalphaxiv

Check my repo out, it’s autoresearch inspired: https://github.com/sparckix/ztare It’s a zero-trust adversarial kernel for validating claims/outputs - with an LLM cheating catalog, deterministic gates, ledgers, and org primitives and a battery of tools to carry out experiments.

Could be useful to harden agent reproductions against gaming or make outputs more auditable. Fully open source.

Happy to share specifics or discuss if any parts are reusable!!

2/N In this case the objective that is hillclimbed is a 0/1 on whether a paper’s initial claims can be matched. In addition to making previous research codebases reproducible, we also believe a tool like this can make it easy for authors to ensure their codebases are legible in the first place.

3/N As models improve, this testbed can extend to more complex tasks that involve actually iterating and improving upon an existing work. Having a public, observable repository of experiments that research agents carry out is going to be invaluable. A slop-free environment where researchers can see exactly what attempts an agent has made when either trying to reproduce a work or make an improvement.

cool

whoo

@askalphaxiv u might want to check this out

@Karthickhps Hi Karthick, yes you can bring your own compute and agent! Would love to chat more just DMed

@askalphaxiv This is great. Is the environment hosted by you, or can users bring their own infra?

@deedydas This is indeed a good feature. I use NotebookLM to listen to the research papers while I am working or traveling. This would be useful as well.

@deedydas Whose GPUs are running?

That's a great UX suggestion! A one-click button on paper pages to fire up autoresearch—no URL editing required—would make agent-powered setup fixes, minimal reproductions, and replication cost estimates ridiculously seamless. The alphaXiv team is building exactly the kind of frictionless tools that accelerate real science. Love it. 🚀

@askalphaxiv this is sick, congrats on launching this!

@askalphaxiv @Grok or just have a button to click on the page that sends people to that too

@askalphaxiv Breaking:

@askalphaxiv @omarshehab FYI