Reinforcement Learning Enables Scaling Generalist Robot Policy Improvement
Under review at ICLR 2027
About
I'm a Master's student at ETH Zurich. I study how reinforcement learning can turn pretrained models into large generalist policies, with a focus on scalable RL, offline RL, and RL post-training for vision-language-action models (VLAs).
I'm always happy to talk about research and collaborations, so feel free to reach out.
- Reinforcement learning
- Robot learning
- Representation learning
Publications
2026
Three Steps at a Time: Learning Representations from Action Sequences in Contrastive RL
NeurIPS 2026 Workshop: From Pretrained Representations to Acting Agents
Under review at ICLR 2027
When Does Non-Uniform Replay Matter in Reinforcement Learning?
NeurIPS 2026 Spotlight
2025
Fair Indivisible Payoffs through Shapley Value
arXiv preprint, 2025
* equal contribution
News
- New preprint: RL Enables Scaling Generalist Robot Policy Improvement.
- Non-Uniform Replay accepted at NeurIPS 2026 as a spotlight. Also named a NeurIPS 2026 Top Reviewer.
- Three Steps at a Time accepted at the NeurIPS 2026 Workshop: From Pretrained Representations to Acting Agents.