PhysEvo: Astra Can Act, Let It

Hugging Face Daily Papers Papers
robotics recursive-self-improvement manipulation frozen-model simulation-to-real agents

Summary

PhysEvo 提出一个围绕冻结模型的物理递归自我改进（RSI）框架，通过元智能体诊断失败并修订工具与技能，在 RoboDojo 任务上取得 62% 成功率，并在真实世界 AgileX PiPER 机器人任务上达到 84% 成功率。

Astra can act, yet reliable manipulation depends on the system through which it observes and controls the world. We introduce PhysEvo, a framework for physical recursive self-improvement (RSI) around a single frozen model. A task agent executes robot tasks; a meta-agent uses the resulting trajectories to diagnose failures, revise tools and skills, and test corrections. The meta-agent can also improve its own diagnostic tools, so retained revisions support both later action and later self-improvement. This process develops joint-level control, evidence-seeking observation, and reusable manipulation skills without model-weight updates or a separately trained action policy. Across 42 RoboDojo tasks, held-out-layout evaluation of retained task-specific deployment versions yields a five-dimension average score of 68.14/100 and 62.00% success, compared with 47.17% for RoboDawn's one-shot Astra agent, the strongest published reference in our comparison. On eight manipulation tasks challenging direct Astra, PhysEvo achieves 55.00% success, compared with 1.25% for the direct-Astra reference. Deploying the simulation-evolved harness on AgileX PiPER and continuing skill revision yields 90.60/100 average score and 84.00% success across 25 trials on five real-world tasks. PhysEvo turns the consequences of action into persistent, testable changes to how a frozen model acts and improves.
Original Article
View Cached Full Text

Cached at: 10/08/26, 12:14 PM

Paper page - PhysEvo: Astra Can Act, Let It

Source: https://huggingface.co/papers/2610.08995

Abstract

Astracanact,yetreliablemanipulationdependsonthesystemthroughwhichitobservesandcontrolstheworld.WeintroducePhysEvo,aframeworkforphysicalrecursiveself-improvement(RSI)aroundasinglefrozenmodel.Ataskagentexecutesrobottasks;ameta-agentusestheresultingtrajectoriestodiagnosefailures,revisetoolsandskills,andtestcorrections.Themeta-agentcanalsoimproveitsowndiagnostictools,soretainedrevisionssupportbothlateractionandlaterself-improvement.Thisprocessdevelopsjoint-levelcontrol,evidence-seekingobservation,andreusablemanipulationskillswithoutmodel-weightupdatesoraseparatelytrainedactionpolicy.Across42RoboDojotasks,held-out-layoutevaluationofretainedtask-specificdeploymentversionsyieldsafive-dimensionaveragescoreof68.14/100and62.00%success,comparedwith47.17%forRoboDawn’sone-shotAstraagent,thestrongestpublishedreferenceinourcomparison.OneightmanipulationtaskschallengingdirectAstra,PhysEvoachieves55.00%success,comparedwith1.25%forthedirect-Astrareference.Deployingthesimulation-evolvedharnessonAgileXPiPERandcontinuingskillrevisionyields90.60/100averagescoreand84.00%successacross25trialsonfivereal-worldtasks.PhysEvoturnstheconsequencesofactionintopersistent,testablechangestohowafrozenmodelactsandimproves.

View arXiv pageView PDFAdd to collection

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2610.08995 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2610.08995 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.08995 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

RobotUse: Allocating Computation, Context, and Decisions

Hugging Face Daily Papers

RobotUse 引入一种围绕指定与修订物理动作来组织计算、上下文与决策的机器人智能体 harness，使智能体通过视觉选择目标与位姿、后端处理几何与运动规划，并在执行中持续学习，最终在 RoboLab 上取得 45% 的任务成功率，超过 CaP-X 6.7 个百分点。

Embodied Turing Machines: Stateful Code for Robot Recursive Self-Improvement

Hugging Face Daily Papers

本文提出 Code-Only-As-Policy (COAP) 范式，将机器人本体视为图灵机，用显式代码替代 VLM/VLA 进行决策与状态跟踪，并借助编码智能体实现闭环递归自我改进；在 RoboDojo 的 42 项双臂任务上达到 70.24% 成功率，且测试时无需模型在环。