Agents Comparison Hub
270 papers - avg viability 5.1
Current research in autonomous agents is increasingly focused on enhancing their efficiency and reliability across various applications, particularly in complex environments. Recent work on multi-agent systems emphasizes the importance of diversity, revealing that heterogeneous configurations can outperform homogeneous ones, thereby addressing diminishing returns in scaling. Innovations like selective memory mechanisms and boundary-aware policy optimization are being developed to improve computational efficiency and decision-making reliability, crucial for real-world applications. Additionally, frameworks such as AgentForge are streamlining the construction of agents by promoting modularity and reducing development time, making it easier for researchers and practitioners to deploy effective solutions. The introduction of skill-based libraries and automated data mining techniques is also addressing the challenges of data quality and richness, which are vital for training effective agents. Collectively, these advancements signal a maturation of the field, paving the way for more robust and adaptable agentic systems capable of tackling intricate tasks in dynamic settings.
Top Papers
- Understanding Agent Scaling in LLM-Based Multi-Agent Systems via Diversity(8.0)
Develop an efficient multi-agent system using diverse LLMs for improved task performance over homogeneous scaling.
- SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks(8.0)
SkillsBench evaluates the effectiveness of procedural Skills in boosting LLM agent task performance.
- A Lightweight Modular Framework for Constructing Autonomous Agents Driven by Large Language Models: Design, Implementation, and Applications in AgentForge(8.0)
AgentForge is an open-source Python framework simplifying the creation and deployment of LLM-driven autonomous agents.
- Avenir-Web: Human-Experience-Imitating Multimodal Web Agents with Mixture of Grounding Experts(8.0)
Avenir-Web: An open-source state-of-the-art agent for executing tasks on dynamic web interfaces using multimodal grounding and adaptive memory.
- Learning to Share: Selective Memory for Efficient Parallel Agentic Systems(8.0)
Launch a system for efficient parallel agentic operations using selective memory to reduce computational cost and enhance task performance.
- M$^2$-Miner: Multi-Agent Enhanced MCTS for Mobile GUI Agent Data Mining(8.0)
Automate high-quality GUI agent data mining with a multi-agent MCTS framework for improved mobile interface interaction.
- CUA-Skill: Develop Skills for Computer Using Agent(8.0)
CUA-Skill provides a structured skills library for autonomous computer-using agents to enhance their efficiency and reliability.
- BAPO: Boundary-Aware Policy Optimization for Reliable Agentic Search(8.0)
Boundary-Aware Policy Optimization enhances reliability for LLM-driven agentic search by teaching AI to recognize its knowledge limits.
- Darwinian Memory: A Training-Free Self-Regulating Memory System for GUI Agent Evolution(8.0)
Develop a training-free, self-evolving memory system for GUI automation that enhances MLLM agents' performance without added costs.
- ProAct: Agentic Lookahead in Interactive Environments(8.0)
ProAct enables AI agents to excel in long-horizon planning with enhanced lookahead reasoning and stable decision-making.