Get hands-on with automated AI research. Try reproducing real research papers using AI agents.
๐๏ธ Nov 30 โ Dec 13
See you all at the hackathon!๐ฅ
Hey, thanks for flagging this! Looks like there are two things going on here. First, you haven't joined the hackathon org on HF yet. Second, we're planning to make the paper claiming Space public closer to the hackathon start date.
That said, we should definitely make the messaging clearer on our end. Thanks for bringing this up!
Get hands-on with automated AI research. Try reproducing real research papers using AI agents.
๐๏ธ Nov 30 โ Dec 13
See you all at the hackathon!๐ฅ
Open models on @huggingface now run through OpenRouter.
Each request goes to the provider with the best price, speed, and uptime for that model, with automatic failover. Everything runs on the same Hugging Face token and API.
You can also send your traces to Hugging Face Storage with Broadcast, so your requests live on the Hub next to your models, ready for things like evals or fine-tuning.
Too many people treat AI as a zero-sum race. Open-source is winning by doing the opposite: collaboration over competition!
This is why weโre excited to release today the first part of our partnership with @OpenRouter to make open models easier to use for all! Hugging Face for storage and collaboration. OpenRouter for inference. Connected so you can:
๐ค Run OpenRouter inference directly on Hugging Face
๐ Access Open model weights directly from OpenRouter
๐ค Store your OpenRouter agent traces on HF to inspect them, train better models or share them with the world!
More collaboration. Fewer zero-sum games. More power to AI builders. Letโs build! ๐๐๐
That's a great question! @asmirkn
$20 can actually go a long way with Inference Providers and HF Jobs, though we understand it won't be enough to fully reproduce some of the more compute-heavy papers that require thousands of GPU hours.
For those cases, the credits can serve as a starting budget, and participants can also aim for a smaller proof of concept rather than a full reproduction. We just ask that any reductions in data, training steps, or seeds are clearly documented, as outlined in the hackathon rules.
Also, an "inconclusive" result is completely valid! Understanding where reproductions fall short, what compromises were needed, and how those affect the results is actually a big part of what we're hoping to learn from this hackathon.
Can AI agents reproduce NeurIPS 2026 papers? Letโs find out.
Pick a paper, give its main claim to ML Intern (or any agent), run the experiments on @huggingface, and report the results.
๐Nov 20 โ Dec 5 ยท Register by Nov 19
๐ณ$20 HF credits for reg. participants. Great prizes!
@Gradio@huggingface Genius! Agents reproducing NeurIPS papers is the perfect real-world eval. Bring the results to Sydney and stop by Turingโs Booth #88.
Can AI agents reproduce NeurIPS 2026 papers? Letโs find out.
Pick a paper, give its main claim to ML Intern (or any agent), run the experiments on @huggingface, and report the results.
๐Nov 20 โ Dec 5 ยท Register by Nov 19
๐ณ$20 HF credits for reg. participants. Great prizes!
๐ Gradio 6.32.0 is out! ๐
New stuff
๐ MCP server now requires login when your app uses auth, and queued MCP calls work with auth
๐งฉ Install gradio CLI extensions straight from GitHub
๐ชช Run history buckets now use the visitorโs own HF token
๐ฅ๏ธ Console forwarding for embedded apps
๐ผ๏ธ Pillow 13 support on 6.x (32-bit integer PNGs still work)
๐ Workflows: 3D model nodes, depth estimation, and masked textboxes
Fixes
๐ท Webcam shows an error when camera permission is denied
๐ช Two Windows fixes: reused cached files arenโt deleted early anymore, and the reserved-port error is clearer
โฑ๏ธ Deferred always_last events now run with their own event data