Reinforcement learning (RL) is a powerful paradigm for policy optimisation in sequential decision making, it is used to train an intelligent agent (policy) to make optimal decisions by maximising some long-term reward. While online RL learns a policy by actively interacting with an environment and collecting new data through exploration and exploitation, offline RL learns a policy from a fixed, pre-collected dataset without further interaction with the environment. A typical challenge in offline RL is distributional shift, where the learned policy may encounter scenarios not covered by the offline data. This paper tackles distributional shift and integration of labelled (with rewards) data and unlabelled (without rewards) data in offline RL.

The proposed method, called semi-pessimistic pseudo labelling (SPL), makes it possible to take advantage of the unlabelled data to find a better policy than would be possible only using the labelled data. This improves the reliability of decision-making by improving estimation and accounting for uncertainty while remaining computationally efficient. The authors provide theoretical guarantees for the method and demonstrate its effectiveness through simulations, standard reinforcement learning benchmarks, and an application to adaptive deep brain stimulation for Parkinson's disease. By making better use of available data, the approach has the potential to improve AI systems in healthcare and other high-stakes domains.

You can find the full ‘Semi-pessimistic Reinforcement Learning’ paper here.

Jin Zhu^1,2, Xin Zhou³, Jiaang Yao⁴, Gholamali Aminian⁵, Omar Rivasplata⁶, Simon Little⁴, Lexin Li³∗, Chengchun Shi¹∗
¹London School of Economics and Political Science, ²University of Birmingham, ³University of California at Berkeley, ⁴University of California at San Francesco, ⁵The Alan Turing Institute, ⁶University of Manchester. *Corresponding authors.

The University of Manchester has launched a new responsible AI competition in partnership with the Jeremy Coller Foundation, offering a £100,000 equity investment to support one promising early-stage venture from its global community of students, staff and alumni.

Applications are now open for the Coller AI Competition, which is designed to help innovators develop ventures with real-world potential – ensuring technology empowers rather than excludes. The competition will support solutions that use AI in ways that are fair, transparent, sustainable and focused on social benefit. The winner will benefit from the opportunity to work closely with the Jeremy Coller Foundation team to develop their venture.

The initiative forms part of the University’s Challenge Accepted campaign, which champions ideas with the potential to tackle major global challenges.

Backed by Manchester alumnus and philanthropist Jeremy Coller, the competition brings together the University’s research strength, entrepreneurial talent and commitment to responsible innovation. It aims to identify and support founders developing AI solutions that are not only commercially promising but designed to benefit people and society.

Finalists will pitch their ventures at the University’s Innovation Festival in October, where the winner will be selected to receive the £100,000 investment.

Professor Aline Miller, Associate Vice President (Enterprise) at The University of Manchester, said: “This is an exciting opportunity to back one early-stage venture with the support and investment needed to develop responsible AI with real-world potential. We’re looking for founders who want to build something purposeful, practical and capable of making a positive difference.”

Jeremy Coller said, “The AI revolution is already changing and shaping our world at a dizzying speed. Used smartly and responsibly AI can be the key that unlocks solutions to some of the greatest challenges facing this planet, and the Coller AI Competition will help Manchester students, staff and alumni turn that potential into reality.”

The Coller Foundation is the philanthropic vehicle of private equity entrepreneur Jeremy Coller. The Foundation works to create positive impact on globally significant issues with a focus on education, entrepreneurship, pensions innovation and the transition to a more sustainable food system. By supporting ambitious ideas, the Foundation helps drive lasting change and empower the next generation of innovators and problem-solvers.

Application timeline – 2026

The competition is open for applications from any students, staff or alumni from across the University who are working on an AI-driven business. At least one founder in each team must be within one of these categories.

• 14^th July – applications open
• 1^st September – applications close
• 21^st September – finalists notified
• 9^th October – final pitches and award

Find out more and apply now – https://unit-m.co.uk/coller-ai-2026

Join us for a week-long celebration of innovation, from Monday, 5 to Friday, 9 October 2026.