AI & ML interests

None defined yet.

Recent Activity

Organization Card
🤗 Hosted by Gradio and Hugging Face · Sponsors TBD
Verify

NeurIPS 2026 papers

with ML Intern 🔬

Pick a NeurIPS 2026 paper and reproduce its main result with ML Intern. Share your code, logs and verdict so anyone can check your work. Then, if you like, build something on top of it in one of the optional tracks.

✓ VERIFIED ✗ FALSIFIED ? INCONCLUSIVE
Dates
Nov 20 – Dec 5, 2026
Registration
Open now · closes Nov 19
Credits
$20 HF credits to the first 1000 registered participants
Limit
5 papers per person
Tracks
3, optional
Prizes
TBD
The core challenge

Everyone does this: reproduce a paper's main result

For each paper, we pick its central table or figure. That is the paper's claim, for example “our model gets 94.1% accuracy (Table 2)”. Your job is to reproduce it. Every participant does this:

  1. Choose up to 5 papers from our claim list.
  2. Give the paper and its claim to ML Intern or another AI agent.
  3. Run the experiments and compare your result with the paper.
  4. Report your verdict and submit your work.

Your verdict for each claim is one of: verified (you got the same result), falsified (the claim does not hold), or inconclusive (you could not tell).

Optional tracks

Go further: three tracks

Tracks are optional extras on top of the core challenge. Each one asks for something creative or original, built on a paper you reproduced. Each track has its own award.

Track

🏗️ RL environment

Build an RL environment from a paper that lets AI agents practise reproducing its results, or be evaluated on it. Format: Harbor or OpenEnv. Details TBD

Track

🧩 Workflow app

Build a Gradio workflow app that compares the paper with earlier work side by side, or connects it to another related model to make something new.

Track

📊 Explainer

Build a visualization or interactive Space that explains the paper to a general audience.

Can you enter more than one track? Yes.

How to take part

Four steps

1

Join and register

Click “Request to join this org” at the top of this page. Then fill in the registration form. Registration closes on Nov 19.

2

Choose papers

Pick up to 5 papers from our claim list. Each paper comes with its claim.

3

Reproduce

Run the experiments. Keep your code, outputs and logs. Optional: build a track entry on top.

4

Submit

Put your work in a public Space and send it through the submission form TBD. You can edit and resubmit until the deadline.

Your research stack

Do the research on Hugging Face

The goal of the hackathon is to do real AI research on Hugging Face infrastructure: your agent runs the experiments, keeps track of them, and publishes what it makes, all on the Hub.

🧰 First, give your agent the Hugging Face CLI. Install the hf CLI and its agent Skill so your agent knows how to work with everything below. Run hf skills add --global to add the Skill for Claude Code, Codex, Cursor and other agents. Set up the CLI for agents
Compute

⚙️ Jobs

Run training and evaluation scripts on CPUs and GPUs in the cloud. Jobs docs

Work in progress

🪣 Buckets

Keep checkpoints, logs and other intermediate files while you work. Buckets docs

Final artifacts

📦 Model & dataset repos

Publish the models you train and the datasets you build, each with a card that explains it. Repositories docs

Experiment tracking

📈 Trackio

Log the metrics of every run and compare runs in a dashboard. Trackio on GitHub

Models

🔌 Inference Providers

Call open models through one API, for example as baselines, judges or to generate data. Inference Providers docs

Demos

🚀 Spaces

Show your results in an app or a page that anyone can open. Your submission goes in a public Space too. Browse Spaces

The claim list

One paper, one central claim

We are preparing a list of NeurIPS 2026 papers that have released code or model weights. For each paper, we pick its central table or figure as the claim, and our team checks every claim before the hackathon starts. Each claim has an ID (for example neurips26-4821) and points to a table or figure in the paper.

  • Why one list? Everyone works on clear, checked claims, and we can compare your result with other people who reproduced the same paper.
  • Papers without a central result, such as pure theory papers, are not on the list.
What to submit

One public Space per paper

For each paper, your Space must contain:

ItemWhat it means
CodeOne script that runs the experiment from start to finish.
EnvironmentThe exact software versions you used (for example requirements.txt, uv.lock or a Dockerfile).
ResultsThe number the paper reports and the number you got.
LogsThe full output of each run, including runs that failed.
VerdictVerified, falsified or inconclusive, with a short reason.
Agent traceThe log of your AI agent session, saved to a Hub dataset. ML Intern will have a “Publish Trace” button for this Coming soon

Then fill in the submission form. It is a Gradio app that checks your Space, shows a checklist of what is complete and what is missing, and marks your submission as complete. AI agents can submit through the same form, because it is also an MCP server. The form also asks which agent and model you used, how much you helped the agent, and whether you looked at other people's reproductions.

📘 The exact folder layout, a starter template and a participant guide will be shared before the start TBD
Rules

Keep it open and honest

  • Up to 5 papers per person, all from the claim list.
  • Any AI agent is allowed. ML Intern is the main focus of the hackathon. Claude Code, Codex, Cursor and others are welcome if you follow the same public upload rules.
  • Everything is public: code, environment, results, logs and agent trace.
  • Be honest about scale. If you trained for fewer steps or used less data than the paper, say so.
  • Falsified claims need proof: a script that shows the problem, and what kind of problem it is (for example an error in the paper or a bug in the official code).
  • Remove personal data from your logs and trace before you publish them. You are responsible for what you publish. We plan to provide a tool that helps with this TBD
  • Changes after the deadline do not count. We judge your Space as it was at your last submission.
  • Teams: 2 to 5 members. · Terms and conditions: TBD
Judging & leaderboard

How entries are judged

  • Numbers: we re-run your code and compare the result with the number in the paper.
  • Figures and plots: scored with a rubric, by AI judges and then by people. TBD
  • Track entries: judged separately for each track. Criteria TBD
  • Leaderboard: a live leaderboard will show submissions as they come in TBD

Full judging criteria: TBD

Awards

Planned awards

🚧 Prizes are still being finalised.

Core challenge

🔬 Best reproduction

Careful, complete and easy to re-run.

Track awards

🏗️ Best RL environment

The most useful environment for agents to practise or be tested on.

🧩 Best workflow app

The most useful or interesting Gradio workflow app built on a paper.

📊 Best explainer

The clearest explanation of a paper for a general audience.

Special prizes

🧨 Best falsification

A claim that does not hold, with experiments that show the actual result and a clear explanation of the problem.

🔌 Best integration

A pull request that adds a paper's model to transformers, diffusers, TRL or another Hugging Face library.

✨ Open artifact

Anything else useful and open that comes out of your reproduction.

🗳️ Community choice

The entry the community likes most.

🤝 Sponsor prizes

TBD

Prize amounts: TBD

Credits & tools

What you get

💳 Hugging Face credits

$20 for each of the first 1000 registered participants, then $10 for everyone who registers after that. How and when you receive them: TBD

🤖 ML Intern

Hugging Face's AI agent for ML work. It can read papers, write code, run jobs and save results to the Hub. Try ML Intern

📮 Submission form

A Gradio app and MCP server that checks your Space and records your submission. Link: TBD

Timeline

Key dates

DateWhat happens
Nov 13 (00:00 GMT)Registration opens
Nov 19Registration closes
Nov 20Hackathon starts
Dec 5Submission deadline
TBDWinners announced

NeurIPS 2026 takes place December 6–13. Plans to share results there: TBD

After the hackathon

Everything stays open

We will publish the claim list, all submissions, track entries, agent traces and verdicts as open datasets on the Hub. Anyone will be able to see which NeurIPS 2026 results were reproduced.

FAQ

Questions

Do I have to use ML Intern?

No. You can use any agent, for example Claude Code, Codex or Cursor. Say what you used in the submission form. ML Intern is the main focus of the hackathon.

What if I cannot reproduce the result?

Submit it anyway. A careful “falsified” or “inconclusive” verdict is a useful result.

Do I have to enter a track?

No. Tracks are optional. Everyone does the core challenge; a track is an extra chance to win an award.

Why only 5 papers?

We want careful work on a few papers, not quick work on many.

Can I change my submission?

Yes, as many times as you like until the deadline.

Can I register after the hackathon starts?

No. Registration closes the day before the start.

What happens to my submission?

All submissions are public and open-sourced, and may be used to build open datasets and benchmarks.

Questions: Discord · @Gradio on X

By taking part you agree to the hackathon terms and conditions TBD, the Hugging Face Code of Conduct and the Content Policy.

models 0

None public yet