Encore: Few-Shot Agentic Discovery of Manipulation Strategies

Hugging Face Daily Papers Papers
robotics manipulation coding-agents few-shot-learning bimanual-manipulation imitation-learning

Summary

ENCORE 提出了一种少样本智能体框架，将机器人操作演示蒸馏为可供编码智能体阅读的证据包，从而自动编写策略程序，在 LIBERO-PRO 上以 96.3% 超过此前最优智能体系统，并在真实双臂机器人上完成魔方交接与杯子翻转任务。

Coding agents can now write, run, and debug programs with little human help. Robot tasks, however, are usually specified by a sentence that leaves out how to grasp, in what order to make contact, and what the result should look like, and an agent given only the sentence must find these details by trial and error. We introduce ENCORE, which gives the agent a few demonstrations as evidence to read rather than as training data. A deterministic builder distills each demonstration into a pack of multi-view keyframes, gripper events, frame strips, and the full trajectory. A coding agent studies the pack, writes a policy program against a fixed perception and action API, refines it iteratively over a few development rollouts, and freezes it before a sealed evaluation that never reveals the success signal. On LIBERO-PRO, the agent's first program already succeeds in half of the perturbed tasks with demonstrations and in one task without them, and the frozen programs outperform the strongest prior agentic system run with the same language model (96.3% against 89.3%). On RoboDojo tasks whose instructions leave the goal unstated, no program succeeds without demonstrations. ENCORE also runs on a real bimanual robot, learning cube handover and cup inversion from five demonstrations each.
Original Article
View Cached Full Text

Cached at: 10/12/26, 10:39 AM

Paper page - Encore: Few-Shot Agentic Discovery of Manipulation Strategies

Source: https://huggingface.co/papers/2609.37359

Abstract

Codingagentscannowwrite,run,anddebugprogramswithlittlehumanhelp.Robottasks,however,areusuallyspecifiedbyasentencethatleavesouthowtograsp,inwhatordertomakecontact,andwhattheresultshouldlooklike,andanagentgivenonlythesentencemustfindthesedetailsbytrialanderror.WeintroduceENCORE,whichgivestheagentafewdemonstrationsasevidencetoreadratherthanastrainingdata.Adeterministicbuilderdistillseachdemonstrationintoapackofmulti-viewkeyframes,gripperevents,framestrips,andthefulltrajectory.Acodingagentstudiesthepack,writesapolicyprogramagainstafixedperceptionandactionAPI,refinesititerativelyoverafewdevelopmentrollouts,andfreezesitbeforeasealedevaluationthatneverrevealsthesuccesssignal.OnLIBERO-PRO,theagent’sfirstprogramalreadysucceedsinhalfoftheperturbedtaskswithdemonstrationsandinonetaskwithoutthem,andthefrozenprogramsoutperformthestrongestprioragenticsystemrunwiththesamelanguagemodel(96.3%against89.3%).OnRoboDojotaskswhoseinstructionsleavethegoalunstated,noprogramsucceedswithoutdemonstrations.ENCOREalsorunsonarealbimanualrobot,learningcubehandoverandcupinversionfromfivedemonstrationseach.

View arXiv pageView PDFProject pageGitHub2Add to collection

Get this paper in your agent:

hf papers read 2609\.37359

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/2609.37359 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.37359 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.37359 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

CoDA: Co-Evolution of Data Engines and Action Policies

Hugging Face Daily Papers

CoDA 提出一种仿真数据引擎与机器人动作策略协同演进的框架，通过共享失败与学习进展记忆、视觉语言诊断以及 Text2potential 分阶段势能奖励，将机器人操作任务的分布外成功率从 5.3% 提升至 27.6%，并在 LIBERO-Pro 上取得显著改善。

OpenRUA: Robot-Use Agents Are Zero-Shot Visuomotor Policies

Hugging Face Daily Papers

OpenRUA 论文提出用现成编码代理（Claude Code、Codex）直接通过原生 ROS 2 控制机器人，作为零样本视觉运动策略，在 CaP-Bench、LIBERO-PRO 和 RoboCasa365 等基准上取得显著的 SOTA 级成绩，并展现出自主编写代码提取 3D 坐标、构建控制库等涌现行为。

HandsOn: A Large-Scale Simulation Benchmark for Generalizable Dexterous Manipulation

Hugging Face Daily Papers

HandsOn 是一个跨具身灵巧操作仿真基准与平台，涵盖 5 种机械臂、6 种灵巧手和约 200 个资产，并提供 18,000 条人类演示与 66 项任务。评测显示当前多任务策略与基于基础模型的控制存在巨大差距，最佳多任务策略在 Core-50 上成功率仅 29.1%。

Skill2Real: Agentic Skill Learning for Zero-Shot Sim-to-Real Robot Manipulation

Hugging Face Daily Papers

Skill2Real introduces an agentic skill learning framework using a Proposer-Verifier-Governor loop that learns executable robot skills in simulation via a shared API and transfers them zero-shot to real-world manipulation without fine-tuning, achieving strong success rates on LIBERO benchmarks and real-robot tasks.