PhysCaP: Grounding Code-as-Policy Agent with Physics-Informed Exploration
Summary
PhysCaP is a physics-informed code-generation agent that actively explores objects to infer hidden physical properties for efficient robotic manipulation.
View Cached Full Text
Cached at: 08/24/26, 08:30 PM
Paper page - PhysCaP: Grounding Code-as-Policy Agent with Physics-Informed Exploration
Source: https://huggingface.co/papers/2608.21031
Abstract
PhysCaP is a physics-informed code-generation agent that actively explores objects to infer hidden physical properties for efficient robotic manipulation.
We present PhysCaP, aPhysics-Informed Code-as-Policyagent for active perception in robotic manipulation. Whilevision-language-action policiesexcel at imitating demonstrations, they rely on passive observation and fail to infer latent physical properties critical for manipulation. PhysCaP augmentscode-as-policyframeworks with a physics-informed exploration layer that enables explicit information-seeking through interaction. It introduces training-freephysical property extractionmodules that estimate object mass and stiffness from robotproprioceptionwithout additional sensors. To balance exploration costs and the efficiency of information obtained, PhysCaP employs adual-agent design: aPlannerthat decides when to explore and when to stop, and aPrioritizerthat filters implausible interactions and ranks the remainder using a heuristic priority score, enabling efficient, targeted exploration. We evaluate PhysCaP on real-world tabletop manipulation tasks (searching for hidden objects, detecting empty cans, and finding ripe avocados) and a simulated task inLIBERO. The results show that existing passive and naive interactive baselines either fail when physical properties are hidden or over-explore, whereas PhysCaP achieves comparable performance with fewer interactions and reduced execution time. Ablation studies further validate the effectiveness of the proposedphysical property extractionmodules. Project page: https://physcap.github.io
View arXiv pageView PDFProject pageAdd to collection
Get this paper in your agent:
hf papers read 2608\.21031
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/2608.21031 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2608.21031 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2608.21031 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
PACE-Bench: Benchmarking Physics Adaptation via Code Evolution in Dynamic Environments
PACE-Bench introduces a simulator-grounded benchmark for evaluating self-evolving agents on physics adaptation tasks involving iterative code redesign after environmental mutations, revealing that mechanism redesign is a major bottleneck compared to parameter inference.
Agent as Policy for Robotic Manipulation
This paper introduces Agent as Policy (AGP), a framework that enables general-purpose agents to directly control physical robots for manipulation tasks without task-specific training, achieving high success rates across various real-world scenarios.
Agent as Policy for Robotic Manipulation
A general-purpose agent directly controls a physical robot by interpreting visuals, writing executable programs, and revising actions based on physical feedback across diverse manipulation tasks, achieving high success rates without task-specific training.
Road to solve physics with AI by generating code for succeeding simulations?
Explores using AI to generate code for physics simulations, aiming to advance solving physics problems through automated code generation.
PhysisForcing: Physics Reinforced World Simulator for Robotic Manipulation
PhysisForcing is a training framework that enhances embodied video generation for robotic manipulation by enforcing physical consistency through pixel-level trajectory alignment and semantic-level relational alignment losses in a DiT-based architecture, achieving notable improvements on benchmarks.