RoboFFT: Finetuning generative robot policy via online reinforcement learning with forward process
cs.RO
Submitted: 2026-09-26
Updated: 2026-09-26
Comments: Accepted by CoRL2026. Project Website: https://student-of-holmes.github.io/RoboFFT/
Project page: https://student-of-holmes.github.io/RoboFFT
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Score-Based Generative Modeling through Stochastic Differential Equations
- Denoising Diffusion Implicit Models
- Flow Matching for Generative Modeling
- ManiFlow: A General Robot Manipulation Policy via Consistency Flow Training
- Counterfactual Behavior Cloning: Offline Imitation Learning from Imperfect Human Demonstrations
- Training Diffusion Models with Reinforcement Learning
- Diffusion Policy Policy Optimization
- ReinFlow: Fine-tuning Flow Matching Policy with Online Reinforcement Learning
- Advantage Weighted Matching: Aligning RL with Pretraining in Diffusion Models
- DiffusionNFT: Online Diffusion Reinforcement with Forward Process
- Flow Matching Policy Gradients
- What Matters in Learning from Offline Human Demonstrations for Robot Manipulation
- Continuous control with deep reinforcement learning
- Proximal Policy Optimization Algorithms
- Dexterous Non-Prehensile Manipulation for Ungraspable Object via Extrinsic Dexterity
- Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations
- Reconciling Reality through Simulation: A Real-to-Sim-to-Real Approach for Robust Manipulation
- Imitation Bootstrapped Reinforcement Learning
- Diffusion Policies as an Expressive Policy Class for Offline Reinforcement Learning
- IDQL: Implicit Q-Learning as an Actor-Critic Method with Diffusion Policies
Related papers
- FMT x: An Efficient and Asymptotically Optimal Extension of the Fast Marching Tree for Dynamic Replanning
- MPCFormer: A physics-informed data-driven approach for explainable socially-aware autonomous driving
- RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies
- ARC: A Reasoning Recipe for Robot Foundation Models
- HRDexDB: A 4D Dexterous Grasping Dataset Across Human and Multiple Robot Embodiments
- APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies