全书目录

附录 G：分章资料清单

G.22 第 11 章正文采用的核心来源

2 分钟 748 字 第 78 / 161 个阅读单元
  • 官方选模框架主轴：OpenAI compare/models、Anthropic Choosing the right model、Google Compare Mode / Gemini thinking、Microsoft model choice guide
  • 多维能力评测主轴：HELM、BBH、ToolLLM、ToolEyes、τ-bench、BFCL V4
  • 长上下文与记忆主轴：Anthropic context engineering、Anthropic context management、OpenAI context personalization、Lost in the Middle、NoLiMa、LONGMEM、Gemini 1.5
  • 系统指标与约束优化主轴：OpenAI Models、OpenAI Latency optimization、Anthropic Models overview、Anthropic Prompt Caching、NVIDIA benchmarking、PagedAttention、Orca
  • 多模态与任务世界主轴：GPT-4o System Card、Moshi、Qwen2.5-Omni、SEAMLESSM4T、Gemini Robotics
  • 风险敏感性与稳定性主轴：HELM、Language Models (Mostly) Know What They Know、GPT-4 System Card、Anthropic Responsible Scaling Policy、Toward expert-level medical question answering、Probabilistic medical predictions