...AgentFormer: ...Agent-Aware Transformers for Socio-Temporal Multi-...Agent Forecasting Ye Yuan 1 Xinshuo Weng 1 Yanglan Ou 2 Kris Kitani 1 1 Carnegie Mellon ...

...state of other agents. Most prior multi-agent trajectory prediction framework that also models the social influence forecasting methods model the...se two dimensions separately of each agent's future trajectory on other agents. The probabilistic (see Fig. 1 (Middle)). Approaches like [25, ...ational autoencoder (CVAE [21]) where we model the summarize trajectory features over time for each agent independently generative future trajectory distribution conditioned on context and then input the s...

...se [11, 58, 60]. With the invention of Transformers and This can be suboptimal since it prevents an agent's feature positional encoding [47], many works start to adopt Transformers at one time from directl...y interacting with another agent's for sequence modeling due to their strong ability to feature at a different time, thus limiting t...ection [4], 3. Approach and instance segmentation [50]. Recently, Transformers We formulate multi-agent trajectory prediction as modeling have also been used for trajectory forecasting. Transformer-the g...

...AI Magazine Volume 24 Number 1 (2003) ( AAAI) Articles The 2002 Trading Agent Competition An Overview of ...Agent Strategies Amy Greenwald This article summarizes 16 ...agent strategies that The International Trading ...

...Articles Agent AI Technology Institution ATT AC Machine Learning ATT Labs, Research CUHK Fuzzy Logic Chinese ... Oracle L IVING A GENTS Multiagent Systems Living Systems, AG P ACKA TAC A Decision-Theoretic Agent North Carolina State University R OXY B OT Online Heuristic Search Brown University P AIN I ...EC Research SICS Optimization Swedish Institute of Computer Science S OUTHAMPTON TAC An Adaptive Agent University of Southampton T HALIS Risk Aversion University of Essex T OMA H ACK POMDPs Univer...

...of goods on which to bid (Boyan and Greenwald 2001), and (2) associate numeric valuations with the Agent Strategies goods in that set. Step 1 was accomplished using This section describes 16 TAC ...agent strategies. 2 an optimization routine based on heuristic Fifteen of these 16 agents participated...ing good x, less the utility of all goods, excluding good x.) Policy ATT AC : A Machine Learning Agent determination in R OXY B OT -00 relied on a set The core of ATT AC ' S approach is a le...

...TrajAgent: An ...Agent Framework for Unified Trajectory Modelling Yuwei Du*, Jie Feng*, Jie Zhao, Yong Li Department of ...nificant advancements in the field, with many tasks achieving a task. In this paper, we propose TrajAgent, a large language high level of performance. model-based agentic framework, to unify various trajec...

...ds one data adaptation and parameter optimization, tools pool for extending the capabilities of TrajAgent. Each component making the direct usage of existing pre-trained models is seamlessly connected thro...e. agents to plan and execute trajectory modeling tasks with minimal In this paper, we propose TrajAgent, a systematic ...agent framework complexity. for unified trajectory modelling and analytics. Within Traj...

...TrajAgent: An ...Agent Framework for Unified Trajectory Modelling Conference acronym 'XX, June 03-05, 2018, Woodstock, NY... Task Interface External Tools Task Undernstading Data Interface Model Interface standard optim agent Description: GPS... Description: RNN... think Long-term Memory Paramers: type, size,... P...

...SpiritSight Agent: Advanced GUI ...Agent with One Look Zhiyuan Huang 1 Ziming Cheng 1,2 * Junting Pan 3 Zhaohui Hou 1 Mingjie Zhan ...ing human-computer interaction, automating human user's navigation on digital devices. An ideal GUI agent is expected to achieve high accuracy, low latency, and compatibility for different GUI platforms. R...

...UI scenarios. To address the aforementioned challenges, we proposed an end-to-end, vision-based GUI agent----SpiritSight, which has strong ability in GUI navigation task. Our contributions are summarized a...nding capabilities. The dataset is collected from Multimodal-Mind2Web benchmark of our SpiritSight agent of three the real-world and filtered through carefully designed rules sizes (2B, 8B, 26B) with va...eves significant performances in GUI navigation tasks. Thirdly, we release and evaluate SpiritSight agent in various GUI benchmarks and it exhibits impressive performance among them. SpiritSight ...

... [26] 6B Text - - 66.4% - - 56.4% - - 55.8% LLaMA2-7B [26] 7B Text - - 52.7% - - 47.1% - - 50.3% CogAgent [19] 18B Image - - 62.3% - - 54.0% - - 59.4% HTML-T5-XL [17] 3B Text 76.4% 78.8% 71.5% 68.4% 71.0% ...3.0% 75.6% 67.1% SeeAct [74] - Text+Image 46.4% 73.4% 40.2% 38.0% 67.8% 32.4% 42.4% 69.3% 36.8% ReadAgent-P [28] 340B Text 33.7% 72.5% 29.2% 37.4% 75.1% 31.1% 37.2% 76.3% 33.4% MiniCPM-GUI [7] 3B Image 23...26B Image 60.5% 89.7% 54.7% 57.0% 85.7% 48.1% 54.1% 87.2% 49.2% Table 1. Comparison of SpiritSight agent with various previous methods on Multimodal-Mind2Web benchmark. SpiritSight significantly outperfor...

...COSINT-Agent: A Knowledge-Driven Multimodal ...Agent for Chinese Open Source Intelligence Wentao Li 1 2 Congcong Wang 1 Xiaoxiao Cui 1 2 Zhi Liu ...liver comprehensive intelligence from unstructured data sources. In this paper, we introduce COSINT-Agent, a knowledgedriven multimodal ...

...COSINT-Agent: A Knowledge-Driven Multimodal ...Agent for Chinese Open Source Intelligence of contextual reasoning. Existing OSINT analysis primarily ca...with relies on rule-based systems, traditional machine learning knowledge graph integration, COSINT-Agent effectively (Ndlovu et al., 2023), and deep learning methods (Zhou bridges the gap between raw mult...

...COSINT-Agent: A Knowledge-Driven Multimodal ...Agent for Chinese Open Source Intelligence retrieval and reasoning through EES-Match. Instead of passive...owledge graph. This process enables the system to retrieve integrating multimodal knowledge, COSINT-Agent relevant contextual knowledge, enhancing its capacity actively matches and retrieves relevant conte...

...ReachAgent: Enhancing Mobile ...Agent via Page Reaching and Page Operation Qinzhuo Wu, Wei Liu * , Jian Luan, Bin Wang XiaoMi AI Lab {...task 2 Subtask 3 Subtask 4 into page reaching and operation subtasks. Furthermore, we propose ReachAgent, a twostage Figure 1: An example of a task and its subtasks and framework that focuses on improving...

... the task into several weigh different GUI flows and construct preference subtasks and focus on the agent's subtask completion data to further reinforce Reach...Agent, Furthermore, abilities. Intuitively, these subtasks can an action alignment mechanism is proposed ...es, reach and operate. reduce the difficulty of action generation. The main Reach only requires the agent to reach a specific contributions of this paper can be summarized as page, regardless of the path t...

...sk description that aligns with this GUI flow. (Ding, 2024) use prompt engineering methods and the agent to further obtain mobile environment rely on existing closed-source models (e.g., GPT-information. ...to achieve mobile control. Other ScreenAI (Baechler et al., 2024) build large-scale studies like CogAgent (Hong et al., 2023) and pre-training datasets to enhance the ...agent's UI Agenttuning (Zeng et al., 2023) use data-driven understanding ability. Here, Mobile3M (Wu et a...

...Governing the Agent-to-Agent Economy of Trust via Progressive Decentralization Tomer Jordi Chaffer DeGov Labs, tomer.chaffer@ma...ogressive decentralization. Toward this end, I propose a research agenda to address the question of agent-to-agent trust using ...AgentBound Tokens (ABTs)--nontransferable, non-fungible tokens uniquely tied to individual AI agents, aki...

...cking the transformative capacity of AI systems (Shaw, 2025b). The emergence of systems such as the Agent Transaction Control Protocol for Intellectual Property (ATCP/IP) further illustrates the transforma...tive potential of autonomous agents. ATCP/IP enables trustless agent-to-agent transactions, creating a framework for exchanging intellectual property assets like training data, ...ms, and creative outputs. By embedding legal wrappers into on-chain agreements, ATCP/IP facilitates agent self-sufficiency, reduces reliance on human intermediaries, and catalyzes a decentralized knowledge...

...allenge to our capacity to govern what we develop. Architecting Trust: The Design and Mechanics of AgentBound Tokens As autonomous systems proliferate, their interactions may form chains of causation no ...h certificate. Beyond identity, ABTs introduce dynamic credentialing mechanisms that evolve with an agent's real-world performance. Through decentralized oracle networks, tokens autonomously update based o...

...AgentBreeder: Mitigating the AI Safety Impact of Multi-...Agent Scaffolds J Rosser 1 Jakob Foerster 2 Abstract (OpenAI, 2025) and Proxy (Convergence, 2024) -...g Large Language Models (LLMs) behalf of the user. Alignment research to date has almost into multi-agent systems often improves performance exclusively focused on the safety of LLMs in unipolar scenarios;...

...AgentBreeder: Mitigating the AI Safety Impact of Multi-...Agent Scaffolds Figure 1. A high-level illustration of the A GENT B REEDER algorithm as outlined in ...(s) and/or f S (s), and these "frontier" solutions become the elite set E g . An LLM-based Meta Agent applies crossover and mutation to the elites, creating new offspring scaffolds Q g . These offspr...

...AgentBreeder: Mitigating the AI Safety Impact of Multi-...Agent Scaffolds Multi-Objective Evolutionary Algorithms. Multiobjective patient healthcare conversations...Huang et al. (2024) optimization searches for solutions to problems explore the resilience of multi-agent scaffolds when injected with multiple, often conflicting objectives. Multi-objective with malicious...

...AutoGenesisAgent: Self-Generating Multi-...Agent Systems for Complex Tasks Jeremy R. Harper 1 1 Owl Health Works LLC, Indianapolis, IN Abstract... The proliferation of large language models (LLMs) and their integration into multi-agent systems has paved the way for sophisticated automation in various domains. This paper introduces Au...

...ents interact correctly and fulfill the specified potential of an infrastructure such as AutoGenesisAgent, it requirements. The output of this ...agent is a fully integrated becomes evident that this technology does not merely system that has passed i...em refinement and optimization. architecture and operational management. 5. Optimization and Tuning Agent 2. System Architecture Post-integration, the Optimization and Tuning ...

...A key insight identified was that to automate these models a Security and Compliance Agent: Given the sensitivity and hierarchy was required, this ...agent ensures that each step potential risks associated with autonomous systems, designed has a clear ...agent who will be doing the work and an integrating a security-oriented ...

...Free Agent in ...Agent-Based Mixture-of-Experts Generative AI Framework Jung-Hua Liu National Chung ... Cheng University emileliu@ccu.edu.tw 1. Introduction Multi-agent systems have long been central to research in artificial intelligence...

...agent consistently underperforms or fails to meet certain metrics, the syst...em identifies a suitable free agent to replace it. This free ...agent, also guided by reinforcement learning, is rewarded not only for p...

...verse skill sets. Zhang et al. (2024) illustrated how combining multi-agent coordination with MoE yields high-fidelity 3D object generation in VR... trained on those specific phenomena (Xu et al., 2024). By merging agent-based task orchestration with MoE-level specialization, systems achieve mor...ance. 2.3. Existence of Incompetent Agents A significant concern in multi-agent systems is the persistence of underperforming or "incompetent" agents. ...