Manifold Research Group tackles ambitious, high-impact research problems that traditional institutions overlook—those too engineering-intensive for academia and too exploratory for industry. Inspired by coordinated research models like ARPAs and FROs, we coordinate focused, cross-functional teams and a large asynchronous research contributor pool to systematically pursue and deliver paradigm-shifting science and technology.
Self-Refinement
Self-Refinement is a research project focused on developing mechanisms by which engineered minds autonomously evaluate, reorganize, calibrate, or improve themselves after training.
The project explores methods for enabling models to assess their own behavior, identify weaknesses or uncertainty, modify internal representations or decision processes, and improve performance without relying solely on additional conventional training.
One current direction investigates pairing diffusion models with energy-based models, allowing a generative model to produce trajectories while a learned evaluator assesses them and provides feedback for improvement. More broadly, the project is intended to explore multiple mechanisms for machine self-refinement.
The Role
OS Team members form the core of Manifold Research Group. As an OS Research Fellow, you will contribute to the development and evaluation of new mechanisms for Self-Refinement.
In this role, you will be responsible for:
- Developing mechanisms for models to evaluate, calibrate, reorganize, or improve themselves
- Designing learning architectures that incorporate internal evaluation or feedback
- Investigating methods for uncertainty estimation and confidence calibration
- Exploring structural adaptation, pruning, representation refinement, and related mechanisms
- Designing experiments and ablations to isolate the effects of self-refinement
- Implementing and evaluating new approaches across relevant machine learning environments and tasks
Qualifications
Outstanding research emerges from individuals who can turn broad ideas into precise, testable mechanisms. For this role, we are looking for:
- Strong background in machine learning, deep learning, or related fields
- Experience with generative models, reinforcement learning, representation learning, or related areas
- Familiarity with uncertainty estimation, model calibration, energy-based models, or diffusion models is a plus
- Experience implementing and training models in PyTorch or similar frameworks
- Ability to formulate clear hypotheses and design rigorous experiments
- Interest in understanding and modifying the internal behavior of learned systems
- Comfort working in ambiguous research settings with evolving problem definitions
Expectations
There are a few key expectations and clarifications regarding the OS Research Team:
- Contribute approximately 10 hours per week to ensure meaningful progress and deep engagement with our projects. Flexibility around life commitments is understood; clear, proactive communication helps us support each other.
- Our working language is English, and strong proficiency is required to clearly communicate technical concepts without confusion or misunderstanding.
- This is a volunteer effort; none of us receive compensation of any kind, including monetary payment, academic credit, or other formal incentives. Our commitment is driven entirely by shared passion for impactful research.
More information on OS Research Team expectations is available here.
We look forward to seeing your application, and hopefully working together soon!