We introduce Ego-R1, a novel framework for reasoning over ultra-long (i.e., in days and weeks) egocentric videos, which leverages a structured Chain-of-Tool-Thought (CoTT) process, orchestrated by an Ego-R1 Agent trained via reinforcement learning (RL). Inspired by human problem-solving strategies, CoTT decomposes complex reasoning into modular steps, with the RL agent invoking specific tools, one per step, to iteratively and collaboratively answer sub-questions tackling such tasks as temporal retrieval and multi-modal understanding. We design a two-stage training paradigm involving supervised finetuning (SFT) of a pretrained language model using CoTT data and RL to enable our agent to dynamically propose step-by-step tools for long-range reasoning. To facilitate training, we construct a dataset called Ego-R1 Data, which consists of Ego-CoTT-25K for SFT and Ego-QA-4.4K for RL. Furthermore, our Ego-R1 agent is evaluated on a newly curated week-long video QA benchmark, Ego-R1 Bench, which contains human-verified QA pairs from hybrid sources. Extensive results demonstrate that the dynamic, tool-augmented chain-of-thought reasoning by our Ego-R1 Agent can effectively tackle the unique challenges of understanding ultra-long egocentric videos, significantly extending the time coverage from few hours to a week.
To address the challenge of passivation of magnesium metal anodes within conventional organic electrolytes, the potential of bamboo biochars (BCs) to facilitate Mg-ion transport opens a new avenue for the development of Mg metal batteries. In this work, BCs with a natural hierarchical structure and a three-dimensional (3D) interconnected network are designed and fabricated via one-step carbonization, followed by a pressure infiltration process to prepare a 3D-BCs/Mg composite. This structure enabled a 3D encapsulation and protection of the Mg metal by the BCs framework. Further investigations revealed that the interconnected 3D-BC skeleton significantly increased the number of ion transport channels and improved ion transport efficiency, while effectively mitigating the reductive decomposition of the electrolyte on the Mg metal surface, as compared to randomly distributed particulate BCs (P-BCs). In a Mg||Mg symmetric cell and a Mg||V2O5 full cell, the BCs/Mg composite anode featuring a 3D ion-conductive protective interface exhibited remarkable performance enhancements. Moreover, the pressure infiltration process provides a cost-effective and simplified fabrication strategy for Mg metal anodes, which promotes the scalable utilization of plant-derived materials to develop environmentally friendly electrode systems. This approach also holds great potential for application in other metal battery systems requiring interfacial protection.
We report a comprehensive study of the electronic structure and magnetic properties of SrRuO3 (SRO) nanoparticles synthesized via the co-precipitation method. Synchrotron-based X-ray photoemission and absorption spectroscopies, including resonant photoemission spectroscopy (RPES), X-ray absorption spectroscopy (XAS), and X-ray magnetic circular dichroism (XMCD), reveal the presence of correlated Ru 4d electrons near the Fermi level. Both the valence-band and conduction-band spectra show dominant incoherent spectral weight, indicating the partial localization of Ru-4d states. The incoherent feature is attributed to the O 2p screening of Ru 4d orbitals at 1.9 eV in the valence band. The nanoparticles exhibit robust ferromagnetism at 70 K, with spin and orbital magnetic moments ranging from 1.23 to 2.23 & micro;B per Ru and from 0.09 to 0.14 & micro;B per Ru, respectively, under external magnetic fields of 0.1-3 T. The observed enhancement in the magnetic moments is attributed to oxygen-vacancy-induced localization, consistent with a correlated metallic ground state in the positive charge-transfer-energy regime. These findings highlight the crucial role of non-stoichiometry in tuning the electronic correlations and magnetic behavior of SRO, underscoring its potential for oxide-based spintronic applications.
Action Quality Assessment (AQA)—the ability to quantify the quality of human motion, actions, or skill levels and provide feedback—has far-reaching implications in areas such as low-cost physiotherapy, sports training, and workforce development. As such, it has become a critical field in computer vision and video understanding over the past decade. Significant progress has been made in AQA methodologies, datasets, and applications, yet a pressing need remains for a comprehensive synthesis of this rapidly evolving field. In this paper, we present a thorough survey of the AQA landscape, systematically reviewing over 200 research papers using the preferred reporting items for systematic reviews and meta-analyses (PRISMA) framework. We begin by covering foundational concepts and definitions, then move to general frameworks and performance metrics, and finally discuss the latest advances in methodologies and datasets. This survey provides a detailed analysis of research trends, performance comparisons, challenges, and future directions. Through this work, we aim to offer a valuable resource for both newcomers and experienced researchers, promoting further exploration and progress in AQA.
Large Language Model-based Multi-Agent Systems (LLM-MASs) have demonstrated remarkable real-world capabilities, effectively collaborating to complete complex tasks. While these systems are designed with safety mechanisms, such as rejecting harmful instructions through alignment, their security remains largely unexplored. This gap leaves LLM-MASs vulnerable to targeted disruptions. In this paper, we introduce Contagious Recursive Blocking Attacks (Corba), a novel and simple yet highly effective attack that disrupts interactions between agents within an LLM-MAS. Corba leverages two key properties: its contagious nature allows it to propagate across arbitrary network topologies, while its recursive property enables sustained depletion of computational resources. Notably, these blocking attacks often involve seemingly benign instructions, making them particularly challenging to mitigate using conventional alignment methods. We evaluate Corba on two widely-used LLM-MASs, namely, AutoGen and Camel across various topologies and commercial models. Additionally, we conduct more extensive experiments in open-ended interactive LLM-MASs, demonstrating the effectiveness of Corba in complex topology structures and open-source models. Our code is available at: https://github.com/zhrli324/Corba.
