数学 llm agent github paper workflow llms 工具 developer-tools arxiv mcp memory rag multi-agent 开发 推荐系统 机器学习 语音识别 daily-pick nlp tool-use no-pick reinforcement-learning 搜索 深度学习 coding-agent observability sandbox spark verification 开发语言 AI Agent Multi-Agent bert claude-code daily-paper deer-flow planning retrieval skills 信息抽取 数据结构 Benchmark LLM LLM Agent Tool Use agents ai-coding archon benchmark code-intelligence codex coordination deep-research eval gui-agent hallucination hermes-agent hierarchy inference langchain llm-systems multimodal orchestration orgagent project python safety speculative-decoding token-efficiency