REVIEW 24 cited by
Relay Policy Learning: Solving Long-Horizon Tasks via Imitation and Reinforcement Learning
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Relay Policy Learning: Solving Long-Horizon Tasks via Imitation and Reinforcement Learning
read the original abstract
We present relay policy learning, a method for imitation and reinforcement learning that can solve multi-stage, long-horizon robotic tasks. This general and universally-applicable, two-phase approach consists of an imitation learning stage that produces goal-conditioned hierarchical policies, and a reinforcement learning phase that finetunes these policies for task performance. Our method, while not necessarily perfect at imitation learning, is very amenable to further improvement via environment interaction, allowing it to scale to challenging long-horizon tasks. We simplify the long-horizon policy learning problem by using a novel data-relabeling algorithm for learning goal-conditioned hierarchical policies, where the low-level only acts for a fixed number of steps, regardless of the goal achieved. While we rely on demonstration data to bootstrap policy learning, we do not assume access to demonstrations of every specific tasks that is being solved, and instead leverage unstructured and unsegmented demonstrations of semantically meaningful behaviors that are not only less burdensome to provide, but also can greatly facilitate further improvement using reinforcement learning. We demonstrate the effectiveness of our method on a number of multi-stage, long-horizon manipulation tasks in a challenging kitchen simulation environment. Videos are available at https://relay-policy-learning.github.io/
Forward citations
Cited by 24 Pith papers
-
D4RL: Datasets for Deep Data-Driven Reinforcement Learning
D4RL supplies new offline RL benchmarks and datasets from expert and mixed sources to expose weaknesses in existing algorithms and standardize evaluation.
-
Automating Potential-based Reward Shaping with Vision Language Model Guidance
VLM-PBRS trains a potential function from small-VLM preferences to enable PBRS in RL, improving sample efficiency in Meta-World and Franka Kitchen without reward hacking.
-
Understanding Multimodal Failure in Action-Chunking Behavioral Cloning
The paper identifies distinct failure mechanisms: excessive posterior-prior regularization erases mode information in latent policies, while smooth base-to-action maps limit mode coverage in generative policies.
-
ScoRe-Flow: Complete Distributional Control via Score-Based Reinforcement Learning for Flow Matching
ScoRe-Flow achieves decoupled mean-variance control in stochastic flow matching by deriving a closed-form score for drift modulation plus learned variance, yielding faster RL convergence and higher success rates on lo...
-
VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Training
VIP learns a visual embedding from human videos whose distance defines dense, smooth rewards for arbitrary goal-image robot tasks without task-specific fine-tuning.
-
Learning Task-Sufficient World Models by Synergizing Agentic Exploration and Structured Modeling
Closed-loop agentic probing plus minimality/sufficiency masking recovers compact task-sufficient world-model latents that improve sample-efficient policy learning and cross-task generalization.
-
DiLA: Disentangled Latent Action World Models
DiLA uses content-structure disentanglement driven by predictive bottlenecks to create semantically structured latent actions for high-fidelity video world models.
-
Learning Reactive Dexterous Grasping via Hierarchical Task-Space RL Planning and Joint-Space QP Control
A multi-agent RL high-level planner outputs task-space velocities that a GPU-parallel QP low-level controller converts to joint velocities while enforcing limits and collisions, yielding robust sim-to-real dexterous g...
-
OGPO: Sample Efficient Full-Finetuning of Generative Control Policies
OGPO is a sample-efficient off-policy method for full finetuning of generative control policies that reaches SOTA on robotic manipulation tasks and can recover from poor behavior-cloning initializations without expert data.
-
OGPO: Sample Efficient Full-Finetuning of Generative Control Policies
OGPO enables sample-efficient full-finetuning of generative control policies via off-policy critics and modified PPO, achieving SOTA on robot manipulation tasks while rescuing poorly initialized behavior cloning polic...
-
Efficient Multi-Objective Planning with Weighted Maximization Using Large Neighbourhood Search
A new LNS-based planner solves weighted-max multi-objective problems 10-100x faster than prior methods with comparable solution quality.
-
DASIP: Dynamic Test-Time Compute Scaling for Robot Control with Stochastic Interpolant Policies
A difficulty classifier adaptively selects step count, solver, and ODE/SDE mode for stochastic-interpolant robot policies, reporting 2.6–4.4x compute savings with roughly unchanged success rates.
-
Learning Upper Lower Value Envelopes to Shape Online RL: A Principled Approach
A two-stage RL framework learns value-function envelopes from offline data and uses them to shape online exploration, yielding regret bounds that improve as offline data grows.
-
Block-wise Adaptive Caching for Accelerating Diffusion Policy
BAC accelerates transformer-based Diffusion Policy up to 3x by block-level adaptive feature caching using an Adaptive Caching Scheduler and Bubbling Union Algorithm to control error propagation.
-
Direct Preference Optimization for Primitive-Enabled Hierarchical RL: A Bilevel Approach
DIPPER uses bi-level optimization and DPO to train the higher-level policy from stationary preference comparisons and value regularization, claiming up to 40% gains on robotic navigation and manipulation tasks while i...
-
Diffusion Policy Policy Optimization
DPPO fine-tunes diffusion policies via policy gradients and outperforms prior RL approaches for diffusion policies and PG-tuned alternatives on robot benchmarks while enabling stable training and hardware deployment.
-
Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulation
A GPT-style model pre-trained on large video datasets achieves 94.9% success on CALVIN multi-task manipulation and 85.4% zero-shot generalization, outperforming prior baselines.
-
Relative Value Learning
A critic that learns antisymmetric value differences ∆(s_i,s_j)=V(s_i)−V(s_j) has a provably contracting Bellman operator and an unbiased advantage estimator, and PPO with this critic matches standard PPO on Atari.
-
Dual-Process Atomic Skill Learning: Decoupling Semantic Reasoning and Real-Time Control
Asynchronous dual-frequency hierarchical imitation learning with VQ skills and training-only latent diffusion improves compositional language-conditioned robot control and reduces skill codebook collapse.
-
Drift Flow Matching
Drift Flow Matching connects direct transport maps from Drift Models with flow-based iterative refinement to enable adaptive computation in generative modeling.
-
Learning 3D Representations for Spatial Intelligence from Unposed Multi-View Images
UniSplat learns consistent 3D geometry, appearance, and semantics from unposed images using dual masking, progressive Gaussian splatting, and recalibration to align predictions across tasks.
-
Sparse ActionGen: Accelerating Diffusion Policy with Real-time Pruning
Sparse ActionGen accelerates diffusion policies up to 4x for robot control via rollout-adaptive pruning and zig-zag activation reuse without performance loss.
-
World Action Models: The Next Frontier in Embodied AI
The paper introduces World Action Models as a new paradigm unifying predictive world modeling with action generation in embodied foundation models and provides a taxonomy of existing approaches.
-
Data Pyramid for Embodied Manipulation
Embodied training data form a five-layer pyramid—real-robot, UMI, ego/exo, simulation, general V–L—ordered by the trade-off between scale and robot alignment, and model capabilities track how those layers are mixed.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.