# Generalizing from References using a Multi-Task Reference and Goal-Driven RL Framework

arXiv: https://arxiv.org/abs/2602.20375 · HTML: https://mlsanity.com/paper/2602.20375/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in RL for robot locomotion and manipulation, one of 15 directions in Deep reinforcement learning methods.
- Builds on 31 of its 39 references that are in the map.
- First posted on arXiv 23 Feb 2026.

## Placed in

- Deep reinforcement learning methods / [RL for robot locomotion and manipulation](https://mlsanity.com/deeprl/rl-for-robot-locomotion-and-manipulation/)

## Builds on (most cited first, 30 of 31)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Asymmetric Actor Critic for Image-Based Robot Learning](https://mlsanity.com/paper/1710.06542/) (2017)
- [OmniH2O: Universal and Dexterous Human-to-Humanoid Whole-Body Teleoperation and Learning](https://mlsanity.com/paper/2406.08858/) (2024)
- [BeyondMimic: From Motion Tracking to Versatile Humanoid Control via Guided Diffusion](https://mlsanity.com/paper/2508.08241/) (2025)
- [ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills](https://mlsanity.com/paper/2502.01143/) (2025)
- [MaskedMimic: Unified Physics-Based Character Control Through Masked Motion Inpainting](https://mlsanity.com/paper/2409.14393/) (2024)
- [SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control](https://mlsanity.com/paper/2511.07820/) (2025)
- [Humanoid Parkour Learning](https://mlsanity.com/paper/2406.10759/) (2024)
- [GMT: General Motion Tracking for Humanoid Whole-Body Control](https://mlsanity.com/paper/2506.14770/) (2025)
- [Retargeting Matters: General Motion Retargeting for Humanoid Motion Tracking](https://mlsanity.com/paper/2510.02252/) (2025)
- [Visual Imitation Enables Contextual Humanoid Control](https://mlsanity.com/paper/2505.03729/) (2025)
- [Advancing Humanoid Locomotion: Mastering Challenging Terrains with Denoising World Model Learning](https://mlsanity.com/paper/2408.14472/) (2024)
- [C·ASE: Learning Conditional Adversarial Skill Embeddings for Physics-based Characters](https://mlsanity.com/paper/2309.11351/) (2023)
- [KungfuBot: Physics-Based Humanoid Whole-Body Control for Learning Highly-Dynamic Skills](https://mlsanity.com/paper/2506.12851/) (2025)
- [Mobile-TeleVision: Predictive Motion Priors for Humanoid Whole-Body Control](https://mlsanity.com/paper/2412.07773/) (2024)
- [TokenHSI: Unified Synthesis of Physical Human-Scene Interactions through Task Tokenization](https://mlsanity.com/paper/2503.19901/) (2025)
- [BeamDojo: Learning Agile Humanoid Locomotion on Sparse Footholds](https://mlsanity.com/paper/2502.10363/) (2025)
- [Attention-Based Map Encoding for Learning Generalized Legged Locomotion](https://mlsanity.com/paper/2506.09588/) (2025)
- [HDMI: Learning Interactive Humanoid Whole-Body Control from Human Videos](https://mlsanity.com/paper/2509.16757/) (2025)
- [HumanMimic: Learning Natural Locomotion and Transitions for Humanoid Robot via Wasserstein Adversarial Imitation](https://mlsanity.com/paper/2309.14225/) (2023)
- [PARC: Physics-based Augmentation with Reinforcement Learning for Character Controllers](https://mlsanity.com/paper/2505.04002/) (2025)
- [VisualMimic: Visual Humanoid Loco-Manipulation via Motion Tracking and Generation](https://mlsanity.com/paper/2509.20322/) (2025)
- [Strategy and Skill Learning for Physics-based Table Tennis Animation](https://mlsanity.com/paper/2407.16210/) (2024)
- [Gallant: Voxel Grid-based Humanoid Locomotion and Local-navigation across 3D Constrained Terrains](https://mlsanity.com/paper/2511.14625/) (2025)
- [KungfuBot2: Learning Versatile Motion Skills for Humanoid Whole-Body Control](https://mlsanity.com/paper/2509.16638/) (2025)
- [ZEST: Zero-shot Embodied Skill Transfer for Athletic Robot Control](https://mlsanity.com/paper/2602.00401/) (2026)
- [MoRE: Mixture of Residual Experts for Humanoid Lifelike Gaits Learning on Complex Terrains](https://mlsanity.com/paper/2506.08840/) (2025)
- [Humanoid Goalkeeper: Learning from Position Conditioned Task-Motion Constraints](https://mlsanity.com/paper/2510.18002/) (2025)
- [StyleLoco: Generative Adversarial Distillation for Natural Humanoid Robot Locomotion](https://mlsanity.com/paper/2503.15082/) (2025)
- [HIL: Hybrid Imitation Learning of Diverse Parkour Skills from Videos](https://mlsanity.com/paper/2505.12619/) (2025)

## Built on by (most cited first, 1 of 1)

- [Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation](https://mlsanity.com/paper/2609.38172/) (2026)

## Abstract (arXiv)

Learning agile humanoid behaviors from human motion offers a powerful route to natural, coordinated control, but existing approaches face a persistent trade-off: reference-tracking policies are often brittle outside the demonstration dataset, while purely task-driven Reinforcement Learning (RL) can achieve adaptability at the cost of motion quality. We introduce a unified multi-task RL framework that bridges this gap by treating reference motion as a prior for behavioral shaping rather than a deployment-time constraint. A single goal-conditioned policy is trained jointly on two tasks that share the same observation and action spaces, but differ in their initialization schemes, command spaces, and reward structures: (i) a reference-guided imitation task in which reference trajectories define dense imitation rewards but are not provided as policy inputs, and (ii) a goal-conditioned generalization task in which goals are sampled independently of any reference and where rewards reflect only task success. By co-optimizing these objectives within a shared formulation, the policy acquires structured, human-like motor skills from dense reference supervision while learning to adapt these skills to novel goals and initial conditions. This is achieved without adversarial objectives, explicit trajectory tracking, phase variables, or reference-dependent inference. We evaluate the method on a challenging box-based parkour playground that demands diverse athletic behaviors (e.g., jumping and climbing), and show that the learned controller transfers beyond the reference distribution while preserving motion naturalness. Finally, we demonstrate long-horizon behavior generation by composing multiple learned skills, illustrating the flexibility of the learned polices in complex scenarios.