Latest 15 Papers - May 19, 2025
Agent
SceneGenAgent: Precise Industrial Scene Generation with Coding Agent
- Title: SceneGenAgent: Precise Industrial Scene Generation with Coding Agent
- Date: 2025-05-15
- Comment:
SceneGenAgent is a novel approach to generating precise industrial scenes using a coding agent. The agent is trained on a dataset of industrial scenes and learns to generate new scenes that are similar in style and content. The agent uses a combination of generative adversarial networks (GANs) and reinforcement learning to generate scenes that are both realistic and diverse.
Fixing Incomplete Value Function Decomposition for Multi-Agent Reinforcement Learning
- Title: Fixing Incomplete Value Function Decomposition for Multi-Agent Reinforcement Learning
- Date: 2025-05-15
- Comment:
In multi-agent reinforcement learning, value function decomposition is a crucial step in learning the optimal policy. However, traditional methods often suffer from incomplete value function decomposition, leading to suboptimal policies. This paper proposes a novel approach to fix incomplete value function decomposition by introducing a new decomposition method that takes into account the interactions between agents.
AI Agents vs. Agentic AI: A Conceptual Taxonomy, Applications and Challenge
- Title: AI Agents vs. Agentic AI: A Conceptual Taxonomy, Applications and Challenge
- Date: 2025-05-15
- Comment: 32 pages, 14 figures, 11 tables
This paper presents a conceptual taxonomy of AI agents and agentic AI, highlighting their differences and similarities. The authors discuss the applications and challenges of each type of AI and provide a comprehensive overview of the field.
Learning Graph Representation of Agent Diffusers
- Title: Learning Graph Representation of Agent Diffusers
- Date: 2025-05-15
- Comment:
Accep...
Accepted at AAMAS2025 International Conference on Autonomous Agents and Multiagent Systems
This paper proposes a novel approach to learning graph representation of agent diffusers. The authors use a graph neural network to learn the representation of the agent diffusers and demonstrate its effectiveness in a multi-agent reinforcement learning setting.
Multi-Agent Path Finding For Large Agents Is Intractable
- Title: Multi-Agent Path Finding For Large Agents Is Intractable
- Date: 2025-05-15
- Comment:
This paper shows that multi-agent path finding for large agents is intractable. The authors provide a formal proof of the intractability and discuss the implications for multi-agent reinforcement learning.
Efficient Adaptation of Reinforcement Learning Agents to Sudden Environmental Change
- Title: Efficient Adaptation of Reinforcement Learning Agents to Sudden Environmental Change
- Date: 2025-05-15
- Comment:
PhD D...
PhD Dissertation, 131 pages
This paper proposes a novel approach to efficient adaptation of reinforcement learning agents to sudden environmental change. The authors use a combination of reinforcement learning and transfer learning to adapt the agent to the new environment.
Random Walks Performed by Topologically-Specific Agents on Complex Networks
- Title: Random Walks Performed by Topologically-Specific Agents on Complex Networks
- Date: 2025-05-15
- Comment: 21 pages, 15 figures
This paper studies the behavior of topologically-specific agents performing random walks on complex networks. The authors provide a comprehensive analysis of the behavior of the agents and discuss the implications for network science.
AutoPentest: Enhancing Vulnerability Management With Autonomous LLM Agents
- Title: AutoPentest: Enhancing Vulnerability Management With Autonomous LLM Agents
- Date: 2025-05-15
- Comment:
24 pa...
24 pages, 1 figure, for implementation, see https://github.com/JuliusHenke/autopentest
This paper proposes a novel approach to enhancing vulnerability management using autonomous LLM agents. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
MASS: Multi-Agent Simulation Scaling for Portfolio Construction
- Title: MASS: Multi-Agent Simulation Scaling for Portfolio Construction
- Date: 2025-05-15
- Comment:
This paper proposes a novel approach to multi-agent simulation scaling for portfolio construction. The authors use a combination of multi-agent reinforcement learning and transfer learning to adapt the agent to the new environment.
Learning Progress Driven Multi-Agent Curriculum
- Title: Learning Progress Driven Multi-Agent Curriculum
- Date: 2025-05-15
- Comment: ICML 2025
This paper proposes a novel approach to learning progress driven multi-agent curriculum. The authors use a combination of reinforcement learning and transfer learning to adapt the agent to the new environment.
AriGraph: Learning Knowledge Graph World Models with Episodic Memory for LLM Agents
- Title: AriGraph: Learning Knowledge Graph World Models with Episodic Memory for LLM Agents
- Date: 2025-05-15
- Comment:
Code ...
Code for this work is avaliable at https://github.com/AIRI-Institute/AriGraph
This paper proposes a novel approach to learning knowledge graph world models with episodic memory for LLM agents. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Learning Virtual Machine Scheduling in Cloud Computing through Language Agents
- Title: Learning Virtual Machine Scheduling in Cloud Computing through Language Agents
- Date: 2025-05-15
- Comment:
This paper proposes a novel approach to learning virtual machine scheduling in cloud computing through language agents. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Towards user-centered interactive medical image segmentation in VR with an assistive AI agent
- Title: Towards user-centered interactive medical image segmentation in VR with an assistive AI agent
- Date: 2025-05-15
- Comment:
This paper proposes a novel approach to user-centered interactive medical image segmentation in VR with an assistive AI agent. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Pre-Act: Multi-Step Planning and Reasoning Improves Acting in LLM Agents
- Title: Pre-Act: Multi-Step Planning and Reasoning Improves Acting in LLM Agents
- Date: 2025-05-15
- Comment:
This paper proposes a novel approach to multi-step planning and reasoning improves acting in LLM agents. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Design and Evaluation of Generative Agent-based Platform for Human-Assistant Interaction Research: A Tale of 10 User Studies
- Title: Design and Evaluation of Generative Agent-based Platform for Human-Assistant Interaction Research: A Tale of 10 User Studies
- Date: 2025-05-15
- Comment:
This paper proposes a novel approach to design and evaluation of generative agent-based platform for human-assistant interaction research. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Recommendation System
Scalable Approximate Biclique Counting over Large Bipartite Graphs
- Title: Scalable Approximate Biclique Counting over Large Bipartite Graphs
- Date: 2025-05-15
- Comment:
This paper proposes a novel approach to scalable approximate biclique counting over large bipartite graphs. The authors use a combination of graph neural networks and reinforcement learning to adapt the agent to the new environment.
Do LLMs Memorize Recommendation Datasets? A Preliminary Study on MovieLens-1M
- Title: Do LLMs Memorize Recommendation Datasets? A Preliminary Study on MovieLens-1M
- Date: 2025-05-15
- Comment:
This paper proposes a novel approach to do LLMs memorize recommendation datasets? A preliminary study on MovieLens-1M. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Towards More Efficient, Robust, Instance-adaptive, and Generalizable Sequential Decision making
- Title: Towards More Efficient, Robust, Instance-adaptive, and Generalizable Sequential Decision making
- Date: 2025-05-15
- Comment: Ph.D. Thesis
This paper proposes a novel approach to towards more efficient, robust, instance-adaptive, and generalizable sequential decision making. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
How Students Use AI Feedback Matters: Experimental Evidence on Physics Achievement and Autonomy
- Title: How Students Use AI Feedback Matters: Experimental Evidence on Physics Achievement and Autonomy
- Date: 2025-05-15
- Comment:
Agent
Q: What is SceneGenAgent and how does it work?
A: SceneGenAgent is a novel approach to generating precise industrial scenes using a coding agent. The agent is trained on a dataset of industrial scenes and learns to generate new scenes that are similar in style and content. The agent uses a combination of generative adversarial networks (GANs) and reinforcement learning to generate scenes that are both realistic and diverse.
Q: What is the main contribution of the paper "Fixing Incomplete Value Function Decomposition for Multi-Agent Reinforcement Learning"?
A: The main contribution of the paper is the introduction of a new decomposition method that takes into account the interactions between agents. This method is shown to improve the performance of multi-agent reinforcement learning algorithms.
Q: What is the difference between AI Agents and Agentic AI?
A: AI Agents are software programs that can perform tasks on their own, while Agentic AI refers to a type of AI that can make decisions and take actions on its own, similar to a human agent.
Q: What is the main contribution of the paper "Learning Graph Representation of Agent Diffusers"?
A: The main contribution of the paper is the introduction of a novel approach to learning graph representation of agent diffusers. The authors use a graph neural network to learn the representation of the agent diffusers and demonstrate its effectiveness in a multi-agent reinforcement learning setting.
Q: What is the main contribution of the paper "Multi-Agent Path Finding For Large Agents Is Intractable"?
A: The main contribution of the paper is the formal proof of the intractability of multi-agent path finding for large agents. This result has implications for multi-agent reinforcement learning.
Q: What is the main contribution of the paper "Efficient Adaptation of Reinforcement Learning Agents to Sudden Environmental Change"?
A: The main contribution of the paper is the introduction of a novel approach to efficient adaptation of reinforcement learning agents to sudden environmental change. The authors use a combination of reinforcement learning and transfer learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "Random Walks Performed by Topologically-Specific Agents on Complex Networks"?
A: The main contribution of the paper is the study of the behavior of topologically-specific agents performing random walks on complex networks. The authors provide a comprehensive analysis of the behavior of the agents and discuss the implications for network science.
Q: What is the main contribution of the paper "AutoPentest: Enhancing Vulnerability Management With Autonomous LLM Agents"?
A: The main contribution of the paper is the introduction of a novel approach to enhancing vulnerability management using autonomous LLM agents. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "MASS: Multi-Agent Simulation Scaling for Portfolio Construction"?
A: The main contribution of the paper is the introduction of a novel approach to multi-agent simulation scaling for portfolio construction. The authors use a combination of multi-agent reinforcement learning and transfer learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "Learning Progress Driven Multi-Agent Curriculum"?
A: The main contribution of the paper is the introduction of a novel approach to learning progress driven multi-agent curriculum. The authors use a combination of reinforcement learning and transfer learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "AriGraph: Learning Knowledge Graph World Models with Episodic Memory for LLM Agents"?
A: The main contribution of the paper is the introduction of a novel approach to learning knowledge graph world models with episodic memory for LLM agents. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "Learning Virtual Machine Scheduling in Cloud Computing through Language Agents"?
A: The main contribution of the paper is the introduction of a novel approach to learning virtual machine scheduling in cloud computing through language agents. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "Towards user-centered interactive medical image segmentation in VR with an assistive AI agent"?
A: The main contribution of the paper is the introduction of a novel approach to user-centered interactive medical image segmentation in VR with an assistive AI agent. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "Pre-Act: Multi-Step Planning and Reasoning Improves Acting in LLM Agents"?
A: The main contribution of the paper is the introduction of a novel approach to multi-step planning and reasoning improves acting in LLM agents. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "Design and Evaluation of Generative Agent-based Platform for Human-Assistant Interaction Research: A Tale of 10 User Studies"?
A: The main contribution of the paper is the introduction of a novel approach to design and evaluation of generative agent-based platform for human-assistant interaction research. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Recommendation System
Q: What is the main contribution of the paper "Scalable Approximate Biclique Counting over Large Bipartite Graphs"?
A: The main contribution of the paper is the introduction of a novel approach to scalable approximate biclique counting over large bipartite graphs. The authors use a combination of graph neural networks and reinforcement learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "Do LLMs Memorize Recommendation Datasets? A Preliminary Study on MovieLens-1M"?
A: The main contribution of the paper is the introduction of a novel approach to do LLMs memorize recommendation datasets? A preliminary study on MovieLens-1M. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "Towards More Efficient, Robust, Instance-adaptive, and Generalizable Sequential Decision making"?
A: The main contribution of the paper is the introduction of a novel approach to towards more efficient, robust, instance-adaptive, and generalizable sequential decision making. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "How Students Use AI Feedback Matters: Experimental Evidence on Physics Achievement and Autonomy"?
A: The main contribution of the paper is the introduction of a novel approach to how students use AI feedback matters: experimental evidence on physics achievement and autonomy. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Misinformation Detection
Q: What is the main contribution of the paper "Large Language Models Meet Stance Detection: A Survey of Tasks, Methods, Applications, Challenges and Future Directions"?
A: The main contribution of the paper is the introduction of a comprehensive survey of tasks, methods, applications, challenges, and future directions for large language models in stance detection.
Q: What is the main contribution of the paper "From Millions of Tweets to Actionable Insights: Leveraging LLMs for User Profiling"?
A: The main contribution of the paper is the introduction of a novel approach to leveraging LLMs for user profiling. The authors use a combination of LLMs and reinforcement learning to adapt the agent to the new environment.
Q: What is the main contribution of the paper "Sentiment and Social Signals in the Climate Crisis: A Survey on Analyzing Social Media Responses to Extreme Weather Events"?
A: The main contribution of the paper is the introduction of a comprehensive survey of tasks, methods, applications, challenges, and future directions for analyzing social media responses to extreme weather events.
Q: What is the main contribution of the paper "CAMOUFLAGE: Exploiting Misinformation Detection Systems Through LLM-driven Adversarial Claim Transformation"?
A: The main contribution of the paper is the introduction of a novel approach to exploiting misinformation detection systems through LLM-driven adversarial claim transformation.
Q: What is the main contribution of the paper "Robust Misinformation Detection by Visiting Potential Commonsense Conflict"?
A: The main contribution of the paper is the introduction of a novel approach to robust misinformation detection by visiting potential commonsense conflict.
Q: What is the main contribution of the paper "A Guide to Misinformation Detection Data and Evaluation"?
A: The main contribution of the paper is the introduction of a comprehensive guide to misinformation detection data and evaluation.
Q: What is the main contribution of the paper "Bridging Cognition and Emotion: Empathy-Driven Multimodal Misinformation Detection"?
A: The main contribution of the paper is the introduction of a novel approach to bridging cognition and emotion: empathy-driven multimodal misinformation detection.
Q: What is the main contribution of the paper "Factually: Exploring Wearable Fact-Checking for Augmented Truth Discernment"?
A: The main contribution of the paper is the introduction of a novel approach to wearable fact-checking for augmented truth discernment.
Q: What is the main contribution of the paper "ViClaim: A Multilingual Multilabel Dataset for Automatic Claim Detection in Videos"?
A: The main contribution of the paper is the introduction of a novel multilingual multilabel dataset for automatic claim detection in videos.