Skip to content

Research Institutions and Conferences

Overview

AI Agent research spans multiple interdisciplinary fields, with top global research institutions and academic conferences playing key roles in advancing the field. This section surveys major research institutions, important conferences, and key researchers.

Top Research Institutions

North America

Institution Affiliation Research Focus Representative Work
Stanford HAI Stanford University AI governance, Agent safety Generative Agents
Berkeley BAIR UC Berkeley Foundational research, Robotics RLHF, Gorilla
Princeton NLP Princeton University NLP, Agents SWE-Agent, SWE-bench
CMU LTI Carnegie Mellon Language Technology WebArena
MIT CSAIL MIT AI Systems Multi-agent systems

Industrial Research Labs

Lab Company Research Focus Representative Work
OpenAI Research OpenAI General AI, Agents GPT series, Operator
Anthropic Research Anthropic AI Safety, Alignment Constitutional AI, MCP
Google DeepMind Google Foundational research Gemini, AlphaCode
Meta FAIR Meta Open-source AI research Llama, Toolformer
Microsoft Research Microsoft Multi-agent systems AutoGen, TaskWeaver

China

Institution Research Focus Representative Work
Tsinghua University Agent evaluation, NLP AgentBench, ChatGLM
Peking University Agent frameworks, Reasoning AgentVerse
CAS Institute of Automation Intelligent systems Multi-agent collaboration
Zhejiang University Knowledge graphs, NLP Agent applications
Zhipu AI Large models, Agents GLM series, AutoGLM

Stanford HAI

Stanford Human-Centered AI Institute's contributions to the agent field:

  • Generative Agents (Park et al., 2023): 25 AI agents living in a virtual town
  • HELM: Comprehensive language model evaluation
  • AI Index Report: Annual AI development report
  • AI safety research: Agent safety and alignment

Berkeley BAIR

Berkeley Artificial Intelligence Research:

  • RLHF/DPO: Alignment methods impacting agent training
  • Gorilla: Tool-using LLM
  • LLM Agent Survey: Systematic survey
  • Robotic agents: Agents in the physical world

Important Academic Conferences

Top Conferences

Conference Full Name Frequency Agent Relevance
NeurIPS Neural Information Processing Systems Annual High
ICML International Conference on Machine Learning Annual High
ICLR International Conference on Learning Representations Annual High
AAAI Association for the Advancement of AI Annual High
ACL Association for Computational Linguistics Annual Medium-High
EMNLP Empirical Methods in NLP Annual Medium-High

Agent-Specific Conferences/Workshops

Conference/Workshop Description
AAMAS International Conference on Autonomous Agents and Multi-Agent Systems
Agent Workshop @ NeurIPS Agent-themed workshop at NeurIPS
LLM Agents Workshop Workshop focused on LLM agents
SoLaR @ NeurIPS Socially Responsible Language Modelling

AAMAS

AAMAS is the flagship conference for agents and multi-agent systems:

  • History: Established in 2002, merged from multiple agent-related conferences
  • Scope: Theory, methods, applications
  • Topics: Multi-agent systems, game theory, social simulation, agent communication
  • LLM Agents: Significant increase in LLM agent papers in recent years

NeurIPS/ICML Agent Workshops

Recent agent-related workshops:

Year Workshop Main Topics
2023 Foundation Models for Decision Making LLMs as decision-making agents
2024 LLM Agents Agent architectures, evaluation, safety
2024 Open-World Agents Agents in open-world environments
2025 Agentic AI Frontiers of agent systems

Key Researchers

Agent Architecture and Systems

Researcher Institution Major Contributions
Shunyu Yao Princeton→OpenAI ReAct, Tree-of-Thoughts
Joon Sung Park Stanford Generative Agents
Harrison Chase LangChain LangChain/LangGraph framework
Chi Wang Microsoft Research AutoGen
Xinyun Chen Google DeepMind Agent evaluation

Agent Safety and Alignment

Researcher Institution Major Contributions
Yarin Gal Oxford Uncertainty estimation
Jacob Steinhardt UC Berkeley AI safety
Dan Hendrycks CAIS Safety evaluation benchmarks

Multi-Agent Systems

Researcher Institution Major Contributions
Michael Wooldridge Oxford Multi-agent systems theory
Tuomas Sandholm CMU Game theory, multi-agent
Qian Liu Tsinghua AgentBench

Key Paper Roadmap

2023-2025 Key Papers

graph TD
    A[2023.03 ReAct] --> B[2023.04 Toolformer]
    B --> C[2023.08 Generative Agents]
    C --> D[2023.10 AgentBench]
    D --> E[2024.01 WebVoyager]
    E --> F[2024.04 SWE-Agent]
    F --> G[2024.06 MCP]
    G --> H[2024.10 Claude Computer Use]
    H --> I[2025 Agent Protocols]

    style A fill:#e3f2fd
    style I fill:#e8f5e9

Hot Directions

  1. Agent safety and alignment: Preventing agents from producing harmful behavior
  2. Multi-agent collaboration: Efficient communication and collaboration between agents
  3. Agent evaluation: Standardized evaluation methods and benchmarks
  4. Agent efficiency: Reducing inference cost and latency
  5. Physical world agents: Robotics and embodied intelligence

Open Questions

  • Long-term goal tracking capabilities for agents
  • How agents can continuously learn from experience
  • Emergent behavior in multi-agent systems
  • Formal verification of agent safety
  • Effective collaboration modes between humans and agents

References

  1. Stanford HAI. "AI Index Report 2024." 2024.
  2. AAMAS. "International Conference on Autonomous Agents and Multi-Agent Systems." Annual.
  3. Yao, S., et al. "ReAct: Synergizing Reasoning and Acting in Language Models." ICLR 2023.
  4. Park, J. S., et al. "Generative Agents: Interactive Simulacra of Human Behavior." UIST 2023.

Cross-references: - Key papers → Key Conferences and Papers - Agent survey → What Is an AI Agent


评论 #