Research & References

Foundational Papers

• ReAct: Synergizing Reasoning and Acting in Language Models — Agent 推理+行动范式奠基（ICLR 2023）
• Toolformer: Language Models Can Teach Themselves to Use Tools — 模型自主学会调用工具（NeurIPS 2023）
• Tree of Thoughts: Deliberate Problem Solving with LLMs — 可回溯的树状推理
• Reflexion: Language Agents with Verbal Reinforcement Learning — 用语言反馈自我改进（NeurIPS 2023）
• Generative Agents: Interactive Simulacra of Human Behavior — 记忆+反思的多 Agent 协作（UIST 2023）
• AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation — 微软多 Agent 对话框架
• A Survey on LLM-Based Autonomous Agents — Agent 领域系统综述
• AgentBench: Evaluating LLMs as Agents — Agent 能力统一基准
• A Survey on LLM-Based Agentic Workflows — Agentic Workflow 专述综述
• GEO: Generative Engine Optimization — 生成式引擎优化方法论（Princeton）