Exploring the Limits of Combinatorial Generalization in Embodied Manipulation

Hugging Face Daily Papers Papers
embodied-manipulation vision-language-action robotics generalization benchmark long-horizon-tasks

Summary

本文提出层级子任务执行（HSE）框架与 RoboCombine 基准，研究具身操作中技能重组的组合泛化极限，指出失败源于「起始状态差距」，并提出 PRTI 轻量后训练策略以提升跨子任务边界的执行成功率。

In embodied manipulation, Vision Language Action (VLA) models can execute atomic skills, but long-horizon manipulation requires recombining these skills without demonstrations for every skill combination. The full instruction route treats skill chaining as long-horizon imitation. In a controlled study, full instruction training on limited skill combination sequences reaches 40.0% ordered success on covered combinations but only 7.7% on recombination, suggesting memorization of sequence templates rather than robust skill recombination. We introduce Hierarchical Subtask Execution (HSE), a diagnostic framework that exposes skill recombination through a shared modular interface over the planner, trigger, and policy, decomposing long-horizon failures into separable modes. We introduce RoboCombine, a benchmark for skill recombination without combined skill training trajectories, with metrics for ordered success and transition conversion across subtask boundaries. RoboCombine reveals that atomic competence alone does not ensure ordered execution. Failures concentrate at transitions where downstream subtasks launch from handoff states from preceding subtasks rather than canonical atomic starts. We identify this mismatch as the start state gap. To address it, we introduce Pose Robust Task Initialization (PRTI), a lightweight atomic task post-training strategy that broadens start state coverage beyond canonical starts. Under oracle timed HSE on RoboCombine, PRTI reaches 18.16% ordered success, compared with 4.21% under random perturbation and 9.47% under a robot-reset oracle, without combined skill training trajectories. These results show that combinatorial generalization requires handoff robust atomic skills that convert first stage progress into reliable downstream execution, not merely decomposition into known subtasks. Code: https://github.com/AutoLab-SAI-SJTU/RoboCombine.
Original Article
View Cached Full Text

Cached at: 10/15/26, 02:55 AM

Paper page - Exploring the Limits of Combinatorial Generalization in Embodied Manipulation

Source: https://huggingface.co/papers/2610.15425

Abstract

Inembodiedmanipulation,VisionLanguageAction(VLA)modelscanexecuteatomicskills,butlong-horizonmanipulationrequiresrecombiningtheseskillswithoutdemonstrationsforeveryskillcombination.Thefullinstructionroutetreatsskillchainingaslong-horizonimitation.Inacontrolledstudy,fullinstructiontrainingonlimitedskillcombinationsequencesreaches40.0%orderedsuccessoncoveredcombinationsbutonly7.7%onrecombination,suggestingmemorizationofsequencetemplatesratherthanrobustskillrecombination.WeintroduceHierarchicalSubtaskExecution(HSE),adiagnosticframeworkthatexposesskillrecombinationthroughasharedmodularinterfaceovertheplanner,trigger,andpolicy,decomposinglong-horizonfailuresintoseparablemodes.WeintroduceRoboCombine,abenchmarkforskillrecombinationwithoutcombinedskilltrainingtrajectories,withmetricsfororderedsuccessandtransitionconversionacrosssubtaskboundaries.RoboCombinerevealsthatatomiccompetencealonedoesnotensureorderedexecution.Failuresconcentrateattransitionswheredownstreamsubtaskslaunchfromhandoffstatesfromprecedingsubtasksratherthancanonicalatomicstarts.Weidentifythismismatchasthestartstategap.Toaddressit,weintroducePoseRobustTaskInitialization(PRTI),alightweightatomictaskpost-trainingstrategythatbroadensstartstatecoveragebeyondcanonicalstarts.UnderoracletimedHSEonRoboCombine,PRTIreaches18.16%orderedsuccess,comparedwith4.21%underrandomperturbationand9.47%underarobot-resetoracle,withoutcombinedskilltrainingtrajectories.Theseresultsshowthatcombinatorialgeneralizationrequireshandoffrobustatomicskillsthatconvertfirststageprogressintoreliabledownstreamexecution,notmerelydecompositionintoknownsubtasks.Code:https://github.com/AutoLab-SAI-SJTU/RoboCombine.

View arXiv pageView PDFGitHubAdd to collection

Get this paper in your agent:

hf papers read 2610\.15425

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2610.15425 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2610.15425 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.15425 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

A GPU-Parallel Framework for Heterogeneous Multi-Task Reinforcement Learning

Hugging Face Daily Papers

本文提出 Hebero，一个基于 GPU 并行 Isaac Lab 的异构多任务机器人操作基准，以及演示引导策略优化方法 DGPO（含 IW-ABC），在 40 项任务上以 50 条演示/任务达到 90.1% 状态输入平均成功率，超越 FAMO-ABC 7.8 个百分点，并在真实 Piper 机器人上验证了单一多任务策略的迁移能力。

EMHO: EMbodied Agent Harness Optimization via Experience Traces

Hugging Face Daily Papers

EMHO 提出一种自进化框架，在冻结具身模型的前提下，通过分析执行轨迹与历史经验自动迭代优化 agent harness（规划、上下文与工具使用），并在 EmbodiedBench 的导航与操作任务上一致提升 Qwen 9B/27B 的任务成功率。

Harness^2: Recursive Agent Harnessing for an Open World

Hugging Face Daily Papers

Harness² 提出了一种即插即用的递归智能体框架（Recursive Agent Harnessing），通过任务级与框架级两层递归改进，在无 ground-truth 验证器的开放世界场景中自动优化智能体执行框架，并在多个专业基准上显著提升了 GLM-5 与 Gemini 3.5 Flash 的表现。