aiqbench analysis
EmbodiedMind addresses critical bottlenecks in training embodied foundation models by introducing Rejection Sampling-based Fine-Tuning and a novel Trie-GRPO algorithm that resolves credit assignment issues in long-horizon planning. This approach significantly improves sample efficiency and task balance, achieving state-of-the-art performance across 18 benchmarks without relying on indiscriminate trajectory-level rewards. The method offers a practical pathway for scaling robotic intelligence through hierarchical policy optimization and strategic data curation.
EmbodiedMind: Adaptive Data Curation and Prefix-Tree Reinforcement Learning for Efficient Embodied Intelligence
arXiv:2609.19659v1 Announce Type: new Abstract: Training embodied foundation models typically requires massive-scale datasets and extensive computational resources, yet often suffers from three critical limitations: (1) inefficient sample utilization due to low-informative samples; (2) imbalanced gradient contributions across heterogeneous tasks; and (3) severe credit assignment problem in long-horizon planning, where trajectory-level rewards indiscriminately penalize all tokens. To address these issues, we propose an efficient training paradigm that achieves state-of-the-art average performance through strategic data selection and hierarchical policy optimization. Our approach consists of three synergistic stages. First, Rejection Sampling-based Fine-Tuning (RSFT) filters out low-informative samples to establish robust behavioral priors while preventing distributional collapse. Second, Iterative Rejection GRPO (IR-GRPO) employs task-specific queues stratified by difficulty to keep datasets balanced across reinforcement learning iterations, coupled with a hybrid reward mechanism for precise cross-task feedback. Third, to enhance long-horizon task planning, we introduce Trie-GRPO, a novel reinforcement learning algorithm based on action prefix trees, which enables step-level advantage estimation. This resolves the credit assignment problem by isolating intermediate correct decisions from downstream errors, while effectively balancing exploration efficiency and depth compared to conventional search trees. As a result, EmbodiedMind achieves a state-of-the-art average performance of 70.02% across 18 benchmarks, and significantly outperforms other embodied foundation models in long-horizon task planning accuracy. Our project will be released for reproducibility.
Abstract:Training embodied foundation models typically requires massive-scale datasets and extensive computational resources, yet often suffers from three critical limitations: (1) inefficient sample utilization due to low-informative samples; (2) imbalanced gradient contributions across heterogeneous tasks; and (3) severe credit assignment problem in long-horizon planning, where trajectory-level rewards indiscriminately penalize all tokens. To address these issues, we propose an efficient training paradigm that achieves state-of-the-art average performance through strategic data selection and hierarchical policy optimization. Our approach consists of three synergistic stages. First, Rejection Sampling-based Fine-Tuning (RSFT) filters out low-informative samples to establish robust behavioral priors while preventing distributional collapse. Second, Iterative Rejection GRPO (IR-GRPO) employs task-specific queues stratified by difficulty to keep datasets balanced across reinforcement learning iterations, coupled with a hybrid reward mechanism for precise cross-task feedback. Third, to enhance long-horizon task planning, we introduce Trie-GRPO, a novel reinforcement learning algorithm based on action prefix trees, which enables step-level advantage estimation. This resolves the credit assignment problem by isolating intermediate correct decisions from downstream errors, while effectively balancing exploration efficiency and depth compared to conventional search trees. As a result, EmbodiedMind achieves a state-of-the-art average performance of 70.02% across 18 benchmarks, and significantly outperforms other embodied foundation models in long-horizon task planning accuracy. Our project will be released for reproducibility.
Robotics (cs.RO); Machine Learning (cs.LG)
arXiv:2609.19659 [cs.RO]
(or arXiv:2609.19659v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.19659 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 306 words
Related coverage
Granite 4.2 LLMs: How They're Built
Hugging Face · Aug 25, 2026
TRACE: A Unified Rollout Budget Allocation Framework for Efficient Agentic Reinforcement Learning
arXiv · Jun 10, 2026
StarOR: Synergizing Tree Search and Test-Time Reinforcement Learning for Optimization Modeling
arXiv · Jun 16, 2026
TRACE: A Unified Rollout Budget Allocation Framework for Efficient Agentic Reinforcement Learning
arXiv · Aug 25, 2026