Shane Gu from Google DeepMind is co-organizing the first ICML workshop pushing reinforcement learning to treat measurable real-world outcomes like efficiency, safety, or economic results as primary training signals instead of defaulting to human preference data. The July 10 2026 event in Seoul's Grand Ballroom draws researchers focused on scaling, continual learning, and recursive self-improvement through these grounded signals.
Flying to ICML in Seuol, Korea and will be around July 8-10 (then a few days in Tokyo). Excited to meet researchers looking into RL scaling + continual learning + recursive self-improvement. Also co-organizing the 1st "RL from World Feedback" workshop https://sites.google.com/view/rlxf-icml2026
The workshop explicitly targets methods for folding in heterogeneous and delayed feedback from actual deployments, yet no established pipelines exist yet for turning those messy signals into reliable training loops across robotics or foundation models.
Speakers from Princeton, Stanford, OpenAI, and Google DeepMind will cover invited talks and a closing panel, with over 150 submissions already received for poster and oral slots that keep the focus on deployable paradigms.
Users are excited about the new RL From World Feedback workshop at ICML because it promises progress on grounded signals like safety and efficiency beyond noisy RLHF, with praise for organizers and anticipation for recordings.
No Digg Deeper questions have been answered for this story yet.
Excited to co-organize the 1st "RL from World Feedback" workshop at ICML on July 10, featuring: - @MillionInt : ex VP of RL @ OpenAI - @chelseabfinn : Stanford/PI - Ben Eysenbach: Princeton - @brianzhan1 : Partner @ http://striker.vc - @robertarail : GDM - Jesse Zhang: UW
Come check out our RLxF workshop this Friday at ICML2026! 🇰🇷 We have fantastic speakers and panels on RL from world feedback: moving beyond human preferences toward signals grounded in real-world outcomes.
📍Fri Jul 10, 8 AM–5 PM KST Grand Ballroom 101–102
Some goals for this workshop: 1) promote RL from world-grounded signals (e.g., efficiency, safety, and economic outcomes) beyond noisy RLHF 2) bridge academia and industry. Great to have @MillionInt @brianzhan1 to speak on latest RL startup trends of Silicon Valley + beyond
Excited to co-organize the 1st "RL from World Feedback" workshop at ICML on July 10, featuring: - @MillionInt : ex VP of RL @ OpenAI - @chelseabfinn : Stanford/PI - Ben Eysenbach: Princeton - @brianzhan1 : Partner @ http://striker.vc - @robertarail : GDM - Jesse Zhang: UW
workshop:
Excited to co-organize the 1st "RL from World Feedback" workshop at ICML on July 10, featuring: - @MillionInt : ex VP of RL @ OpenAI - @chelseabfinn : Stanford/PI - Ben Eysenbach: Princeton - @brianzhan1 : Partner @ http://striker.vc - @robertarail : GDM - Jesse Zhang: UW
@MillionInt @brianzhan1 link: https://sites.google.com/view/rlxf-icml2026
Some goals for this workshop: 1) promote RL from world-grounded signals (e.g., efficiency, safety, and economic outcomes) beyond noisy RLHF 2) bridge academia and industry. Great to have @MillionInt @brianzhan1 to speak on latest RL startup trends of Silicon Valley + beyond
@shaneguML @brianzhan1 @MillionInt @chelseabfinn @robertarail Excited to see you at ICML!!
@shaneguML Amazing! Looking forward to catching up 👋
@shaneguML that workshop sounds like a massive undertaking, congrats on organizing it.
@shaneguML this workshop sounds like the perfect space for actual progress in RL.
where do I sign up to follow the recordings?
