@RoundtableSpace: HUGGING FACE JUST AUTOMATED THEIR ENTIRE POST-TRAINING TEAM WITH AN AGENT. It reads papers, runs GPU experiments, itera…
Summary
Hugging Face replaced its post-training team with an autonomous agent that reads papers, runs GPU experiments, and improves models, achieving a 22-point benchmark jump in under 10 hours and beating Codex on HealthBench by 60%.
View Cached Full Text
Cached at: 04/22/26, 08:29 AM
HUGGING FACE JUST AUTOMATED THEIR ENTIRE POST-TRAINING TEAM WITH AN AGENT. It reads papers, runs GPU experiments, iterates, and builds research-backed models autonomously. Pushed a benchmark from 10% to 32% in <10 hrs. Beat Codex on HealthBench by 60%
Similar Articles
@Thom_Wolf: Love this work from Aksel and the post-training team at Hugging Face! Turns out the HF ecosystem (papers, datasets, mod…
Hugging Face’s post-training team demonstrates how the HF ecosystem enables ML agents to autonomously train any AI model to peak performance.
@socialwithaayan: HUGGING FACE JUST OPEN-SOURCED THE ML INTERN EVERY RESEARCHER HAS DREAMED OF No more spending days reading papers and w…
Hugging Face open-sourced ml-intern, an autonomous agent that reads ML papers, discovers datasets, trains models, debugs failures, and ships production-ready models to the Hub, automating the entire post-training workflow.
@nini_incrypto_: Hugging Face automates entire AI training pipeline! Recently, a project called ml-intern has gone viral on GitHub. It's like a 24/7 algorithmic intern that can independently perform post-training of large models. 1. Autonomous research: It will…
The ml-intern project from Hugging Face has gone viral on GitHub, enabling full automation of the entire workflow including paper research, data processing, training script writing, and model training, without human intervention. It significantly improves the performance of small models (such as Qwen3-1.7B), even surpassing Claude Code.
@ClementDelangue: Hugging Face is becoming the platform for agents to use and build AI. Now they can call 1M HF spaces to do everything t…
Hugging Face now lets AI agents invoke 1 million Spaces, turning the hub into a programmable platform where agents can tap any specialized model or app.
@seelffff: > reads papers on arXiv autonomously > finds and checks datasets on HF Hub > writes the training script itself > genera…
Hugging Face open-sourced ml-intern, an autonomous agent that performs the entire ML post-training loop—reading papers, finding datasets, writing scripts, generating data, monitoring training, and uploading weights—achieving significant GPQA improvement with a 1.7B model in 10 hours without human intervention.