# DeformPAM: Data-Efficient Learning for Long-horizon Deformable Object Manipulation via Preference-based Action Alignment

arXiv: https://arxiv.org/abs/2410.11584 · HTML: https://mlsanity.com/paper/2410.11584/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in Reward modeling and RLHF, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 38 of its 45 references that are in the map.
- First posted on arXiv 15 Oct 2024.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Reward modeling and RLHF](https://mlsanity.com/alignment/reward-modeling-and-rlhf/)
- Offline RL and imitation learning from data / [Robot manipulation imitation learning](https://mlsanity.com/offlinerl/robot-manipulation-imitation-learning/)

## Builds on (most cited first, 30 of 38)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Diffusion Policy: Visuomotor Policy Learning via Action Diffusion](https://mlsanity.com/paper/2303.04137/) (2023)
- [Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware](https://mlsanity.com/paper/2304.13705/) (2023)
- [A General Language Assistant as a Laboratory for Alignment](https://mlsanity.com/paper/2112.00861/) (2021)
- [Grounded SAM: Assembling Open-World Models for Diverse Visual Tasks](https://mlsanity.com/paper/2401.14159/) (2024)
- [Diffusion Model Alignment Using Direct Preference Optimization](https://mlsanity.com/paper/2311.12908/) (2023)
- [Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robots](https://mlsanity.com/paper/2402.10329/) (2024)
- [Reward learning from human preferences and demonstrations in Atari](https://mlsanity.com/paper/1811.06521/) (2018)
- [Learning Latent Plans from Play](https://mlsanity.com/paper/1903.01973/) (2019)
- [Smaug: Fixing Failure Modes of Preference Optimisation with DPO-Positive](https://mlsanity.com/paper/2402.13228/) (2024)
- [MimicPlay: Long-Horizon Imitation Learning by Watching Human Play](https://mlsanity.com/paper/2302.12422/) (2023)
- [Learning to Generalize Across Long-Horizon Tasks from Human Demonstrations](https://mlsanity.com/paper/2003.06085/) (2020)
- [From Play to Policy: Conditional Behavior Generation from Uncurated Robot Data](https://mlsanity.com/paper/2210.10047/) (2022)
- [ARGS: Alignment as Reward-Guided Search](https://mlsanity.com/paper/2402.01694/) (2024)
- [IRIS: Implicit Reinforcement without Interaction at Scale for Learning Control from Offline Robot Manipulation Data](https://mlsanity.com/paper/1911.05321/) (2019)
- [FlingBot: The Unreasonable Effectiveness of Dynamic Manipulation for Cloth Unfolding](https://mlsanity.com/paper/2105.03655/) (2021)
- [Latent Plans for Task-Agnostic Offline Reinforcement Learning](https://mlsanity.com/paper/2209.08959/) (2022)
- [Contrastive Preference Learning: Learning from Human Feedback without RL](https://mlsanity.com/paper/2310.13639/) (2023)
- [Neural Task Programming: Learning to Generalize Across Hierarchical Tasks](https://mlsanity.com/paper/1710.01813/) (2017)
- [RoboCook: Long-Horizon Elasto-Plastic Object Manipulation with Diverse Tools](https://mlsanity.com/paper/2306.14447/) (2023)
- [Active Preference-Based Gaussian Process Regression for Reward Learning](https://mlsanity.com/paper/2005.02575/) (2020)
- [TACO: Learning Task Decomposition via Temporal Alignment for Control](https://mlsanity.com/paper/1803.01840/) (2018)
- [AdaptiGraph: Material-Adaptive Graph-Based Neural Dynamics for Robotic Manipulation](https://mlsanity.com/paper/2407.07889/) (2024)
- [SpeedFolding: Learning Efficient Bimanual Folding of Garments](https://mlsanity.com/paper/2208.10552/) (2022)
- [Cloth Funnels: Canonicalized-Alignment for Multi-Purpose Garment Manipulation](https://mlsanity.com/paper/2210.09347/) (2022)
- [On the Limited Generalization Capability of the Implicit Reward Model Induced by Direct Preference Optimization](https://mlsanity.com/paper/2409.03650/) (2024)

## Built on by (most cited first, 7 of 7)

- [From Action Labels to Sets: Rethinking Action Supervision for Imitation Learning from Corrective Feedback](https://mlsanity.com/paper/2502.07645/) (2025)
- [Preference Aligned Visuomotor Diffusion Policies for Deformable Object Manipulation](https://mlsanity.com/paper/2602.09583/) (2026)
- [FoldNet: Learning Generalizable Closed-Loop Policy for Garment Folding via Keypoint-Driven Asset and Demonstration Synthesis](https://mlsanity.com/paper/2505.09109/) (2025)
- [Set-Supervised Diffusion Policy: Learning Action-Chunking Diffusion through Corrections](https://mlsanity.com/paper/2606.01865/) (2026)
- [ReSeFlow: Rectifying SE(3)-Equivariant Policy Learning Flows](https://mlsanity.com/paper/2509.22695/) (2025)
- [Phase-Conditioned Imitation Learning with Autonomous Failure Recovery for Robust Deformable Object Manipulation](https://mlsanity.com/paper/2605.29407/) (2026)
- [Vision-Based 6-DoF Grasp Pose Estimation for Robot Cloth Unfolding](https://mlsanity.com/paper/2609.31452/) (2026)

## Abstract (arXiv)

In recent years, imitation learning has made progress in the field of robotic manipulation. However, it still faces challenges when addressing complex long-horizon tasks with deformable objects, such as high-dimensional state spaces, complex dynamics, and multimodal action distributions. Traditional imitation learning methods often require a large amount of data and encounter distributional shifts and accumulative errors in these tasks. To address these issues, we propose a data-efficient general learning framework (DeformPAM) based on preference learning and reward-guided action selection. DeformPAM decomposes long-horizon tasks into multiple action primitives, utilizes 3D point cloud inputs and diffusion models to model action distributions, and trains an implicit reward model using human preference data. During the inference phase, the reward model scores multiple candidate actions, selecting the optimal action for execution, thereby reducing the occurrence of anomalous actions and improving task completion quality. Experiments conducted on three challenging real-world long-horizon deformable object manipulation tasks demonstrate the effectiveness of this method. Results show that DeformPAM improves both task completion quality and efficiency compared to baseline methods even with limited data. Code and data will be available at https://deform-pam.robotflow.ai.