Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.26848

Where is Physical RSI capability from? An Observational Scaling Law of LLM-as-policy

Zekai Li·Jiayi Yu·Jinlu Zhang·Luyang Niu·Zekun Qi·Jun Zheng·Xuchuan Chen·Dairu Liu·Yintianrun Zhang·He Wang·Li Yi
View original source ↗Read source document ↗Export metadata ↓

Abstract

In this paper, we systematically investigate the mechanisms, scaling behavior, and generalization of physical Recursive self-improvement. We introduce PhysRSI-eval to quantify physical RSI through a five-axis metric: First-Attempt Performance, Efficiency, Success, Budget and Improvement from failure. Different VLMs are evaluated on 18 manipulation tasks under a fixed interaction budget. The analysis of experimental results motivates the Physical Barrel Hypothesis: different capabilities may constrain different parts of improvement. The five-axis metrics also distinguish systems with close public benchmark averages and show that high scores need not come with early or efficient gains. We study observational scaling from two perspectives: predicting the relative performance of VLM agents from their public capabilities, and forecasting later progress from early physical feedback. Together, these analyses ask how quickly and how far an agent can improve. Moreover, we conduct workspace interventions and then analyze what improvement leaves behind. Removing history reduces late-round success, frozen workspaces remain useful on unseen layouts, and three cross-task cases achieve success earlier. The results suggest that scene-specific adaptation and reusable experience can coexist.

Explore related subjects

cs.ROcs.AIcs.CV

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Zekai Li, Jiayi Yu, Jinlu Zhang, Luyang Niu, Zekun Qi, Jun Zheng, Xuchuan Chen, Dairu Liu, Yintianrun Zhang, He Wang, Li Yi. 2026-10-19. Where is Physical RSI capability from? An Observational Scaling Law of LLM-as-policy. https://arxiv.org/abs/2610.26848

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

AnyNav: Visual Neuro-Symbolic Friction Learning for Off-road Navigation

Off-road navigation is critical for a wide range of field robotics applications from planetary exploration to disaster response. However, it remains a longstanding challenge due to unstructured environments and the inherently complex terrain-vehicle interactions. Traditional physics-based methods struggle to accurately capture the nonlinear dynamics underlying these interactions, while purely data-driven approaches often overfit to specific motion patterns, vehicle geometries, or platforms, limiting their generalization in diverse, real-world scenarios. To address these limitations, we introduce AnyNav, a vision-based friction estimation and navigation framework grounded in neuro-symbolic principles. Our approach integrates neural networks for visual perception with symbolic physical models for reasoning about terrain-vehicle dynamics. To enable self-supervised learning in real-world settings, we adopt the imperative learning paradigm, employing bilevel optimization to train the friction network through physics-based optimization. This explicit incorporation of physical reasoning substantially enhances generalization across terrains, vehicle types, and operational conditions. Leveraging the predicted friction coefficients, we further develop a physics-informed navigation system capable of generating physically feasible, time-efficient paths together with corresponding speed profiles. We demonstrate that AnyNav seamlessly transfers from simulation to real-world robotic platforms, exhibiting strong robustness across different four-wheeled vehicles and diverse off-road environments.

cs.RO↗
arXiv2026

LLM-Flock: LLM-Driven Decentralized Multi-Robot Flocking via Influence-Based Consensus with Collision Avoidance

Large Language Models (LLMs) have shown strong capabilities in reasoning and planning, prompting researchers to explore their use as decentralized decision-makers for multi-robot formation control. However, directly applying LLMs to such tasks often leads to unstable behavior, in which robots may collapse to the centroid of their positions or diverge entirely, due to hallucinated reasoning and limited coordination awareness. We propose LLM-Flock, a decentralized multi-robot flocking framework that addresses these challenges through two key mechanisms: (1) an influence-based consensus protocol, where robots iteratively refine individually generated formation plans by adopting those of more connected neighbors, driving the team toward a unified plan without centralized control, and (2) a two-layer collision avoidance strategy that pairs LLM-based urgency reasoning with a programmatic collision checker and re-prompting mechanism. We evaluate LLM-Flock across five state-of-the-art LLMs, including both closed-source (GPT-5, Claude 4.5) and open-source (Llama3.1-405b, Qwen-Max, DeepSeek-R1) models, demonstrating notable improvements in formation stability and convergence over prior LLM-based methods. Real-world experiments with Crazyflie drones further validate the framework's applicability to physical multi-robot systems.

cs.RO↗
arXiv2026

RoboPocket: Improve Robot Policies Instantly with Your Phone

Scaling robot learning is bottlenecked not only by data volume, but also by whether the data targets the policy's failures. Existing robot-free handheld interfaces enable portable in-the-wild data collection, yet remain open-loop: users record demonstrations without feedback from the current policy. Interactive methods such as DAgger provide such feedback by collecting corrective on-policy data, but require deploying imperfect policies on physical robots, making them costly, unsafe, and difficult to scale. We present RoboPocket, a smartphone-based system for robot-free policy iteration. By combining on-device sensing, display, and communication with cloud-side policy inference, RoboPocket provides real-time collection guidance and visualizes the policy's predicted trajectory in augmented reality as AR Visual Foresight. This allows users to identify policy weaknesses and collect targeted corrective data without deploying a physical robot. An asynchronous finetuning pipeline continuously incorporates new corrections and updates the policy within minutes, realizing a DAgger-inspired robot-free correction loop. Across diverse real-world manipulation tasks, we validate RoboPocket as a scalable data engine and show that online corrective data collection improves sample efficiency by up to $2\times$ over scaling offline demonstrations. Project page and videos: https://robo-pocket.github.io.

cs.RO↗