LLM Powered Autonomous Agents
近期研究聚焦于智能代理系统的任务分解与增强方法,主要包括:1)思维链方法(CoT、Tree of Thoughts等)改进推理过程;2)记忆机制(感官记忆、短时/长时记忆等)增强信息处理;3)工具集成方法(MRKL、Toolformer等)扩展外部能力;4)案例研究显示在科学发现和生成代理等领域的应用潜力。这些技术通过结合规划、记忆和工具使用,显著提升了代理系统的复杂任务处理能力。详见综述文章(h
-
Agent System
-
Planning
-
Task Decomposition
-
Chain of thought (CoT; Wei et al. 2022)
-
https://arxiv.org/abs/2201.11903
-
-
Tree of Thoughts (Yao et al. 2023)
-
https://arxiv.org/abs/2305.10601
-
-
LLM+P (Liu et al. 2023)
-
https://arxiv.org/abs/2304.11477
-
-
-
Self-Reflection
-
ReAct (Yao et al. 2023
-
https://arxiv.org/abs/2210.03629
-
-
Reflexion (Shinn & Labash 2023)
-
https://arxiv.org/abs/2303.11366
-
-
Chain of Hindsight (CoH; Liu et al. 2023)
-
https://arxiv.org/abs/2302.02676
-
Algorithm Distillation (AD; Laskin et al. 2023)
-
https://arxiv.org/abs/2210.14215
-
-
-
-
-
Memory
-
Types of Memory
-
Sensory Memory
-
conic memory (visual)
-
echoic memory (auditory)
-
haptic memory (touch)
-
-
Short-term memory
-
Long-term memory
-
Explicit / declarative memory
-
episodic memory (events and experiences)
-
semantic memory (facts and concepts)
-
-
Implicit / procedural memory
-
-
-
Maximum Inner Product Search
-
LSH
-
ANNOY
-
HNSW
-
FAISS
-
ScaNN
-
-
-
Tool use
-
MRKL (Karpas et al. 2022)
-
TALM (Tool Augmented Language Models; Parisi et al. 2022)
-
Toolformer (Schick et al. 2023)
-
HuggingGPT (Shen et al. 2023)
-
API-Bank (Li et al. 2023)
-
-
Case Studies
-
Scientific Discovery Agent
-
Generative Agents Simulation
-
Proof-of-Concept Examples
-
-
https://lilianweng.github.io/posts/2023-06-23-agent/
-
火山引擎开发者社区是火山引擎打造的AI技术生态平台,聚焦Agent与大模型开发,提供豆包系列模型(图像/视频/视觉)、智能分析与会话工具,并配套评测集、动手实验室及行业案例库。社区通过技术沙龙、挑战赛等活动促进开发者成长,新用户可领50万Tokens权益,助力构建智能应用。
更多推荐
所有评论(0)