Thomas Weng
191 posts
@thomas_weng
Robotics Research Scientist at The AI Institute | @CMU_Robotics PhD
🚨New paper alert 🚨 from @rai_inst! arxiv.org/abs/2603.15757 🤖You robot policy is actually better than you think! We find that for a given policy, ALWAYS denoising a single noise vector, which we call a ✨Golden Ticket ✨, leads to consistent performance improvements! 🧵...
Residual RL is a powerful strategy for adapting a pretrained base policy, but it struggles when: ❌ Exploration is uncontrolled ❌ Base policies are stochastic We tackle this in our new RAL paper “Accelerating Residual Reinforcement Learning with Uncertainty Estimation” (1/5)
Excited to share our NeurIPS 2025 paper introducing our video reasoning framework, ROVER (Reasoning Over VidEo Recursively), that improves visual understanding of VLMs in embodied settings. ROVER is a recursive framework that enables the model to maintain a compact attention window at each timestep of the video, without losing global context across the full video. ROVER works by decomposing the video into segments corresponding to each subtask within the full task trajectory. ROVER then generates a separate line of reasoning for each subtask, instead of attempting to reason across the full trajectory. We evaluate on simulated and real-world robotic manipulation tasks from RoboCasa and OpenX Embodiment. Overall, ROVER significantly improves the ability of VLMs to reason about what is happening at each moment during a robot task attempt. rover-vlm.github.io
And we wrapped with a dinner sponsored by The AI Institute theaiinstitute.com. It was a great social event! The workshop also would not have been possible without sponsorship by the RSS Foundation @RoboticsSciSys, Robohouse robohouse.nl, and TRI @ToyotaResearch!
And we wrapped with a dinner sponsored by The AI Institute theaiinstitute.com. It was a great social event! The workshop also would not have been possible without sponsorship by the RSS Foundation @RoboticsSciSys, Robohouse robohouse.nl, and TRI @ToyotaResearch!
@RoboticsSciSys @CosimoDella @ProfKulic @BekrisKostas Shout out to our OC members who were onsite at Delft! @robobzbz, Manuel Keppler, @thepythagorian, @yan_youcan81764, Rachel Holladay, @MashaItkina, and @thomas_weng We are grateful to all who made this workshop possible, both remote and in-person. Thanks for all your hard work!🙏
Incredibly proud to have hosted the Workshop on Symmetries in Robot Learning at #RSS2023! Insightful talks and engaging discussions. Huge thank to our speakers @tesssmidt @haosu_twitr @GhaffariMaani @andyzeng_ @erikjbekkers @HughWang19 and every participants who have joined us!
And here is our amazing cohort of Pioneers! #RSS2023 #RSSPioneers2023
