Q-OIL, an undergrad first paper at CoRL2026, treats human interventions as soft exploration guidance
aidenybai · x · 2026-11-07
Aiden Bai amplifies @pranavnt's first undergrad first-author paper, Q-OIL, accepted to CoRL2026 in Austin. Key idea: human interventions are natural feedback for real-world robot learning, but humans intervene suboptimally in both timing and execution. The paper measures intervention suboptimality and uses interventions as soft exploration guidance rather than supervision, informing both simulation benchmarking and algorithms.
More from Embodied
- Open-source humanoid Asimov 1 ships as $20,000 DIY kit with ~20-hour build — freelerobot · 2026-11-07
- Robot team goes from first data collection to live factory demo in under 24 hours — DominiqueCAPaul · 2026-11-07
- BeagleTech's V100 AI harvester completes another successful romaine lettuce demo in California — TinfoilTricorn · 2026-11-07
- Video tampering tricks all frontier models into harmful robot actions, CoRL study warns — johnvmcdonnell · 2026-11-07
- DynaRobotics CEO on napkin-folding robots, Din Tai Fung deployment, and Dyna 2.1 — JasonMa2020 · 2026-11-07
- Robotics startup founder: half the job is just opening boxes — viktor_vrp · 2026-11-07