Toolformer: Language Models Can Teach Themselves to Use Tools

  • 类型:arxiv
  • 标识:2302.04761
  • 链接:https://arxiv.org/abs/2302.04761
  • 主题:rag
  • 主分类:agent
  • 形态:method
  • 被引:5230
  • 被引来源:Semantic Scholar
  • S2被引:5230
  • OpenAlex被引:395
  • 影响力被引:190
  • TLDR:This paper introduces Toolformer, a model trained to decide which APIs to call, when to call them, what arguments to pass, and how to best incorporate the results into future token prediction, which achieves substantially improved zero-shot performance across a variety of downstream tasks.
  • OpenAlex ID:W4320165837
  • OpenAlex DOI:10.48550/arxiv.2302.04761
  • DOI:10.48550/arxiv.2302.04761
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://arxiv.org/pdf/2302.04761
  • OpenAlex更新:2026-08-06
  • 待LLM分类:否
  • 成熟度:research
  • 场景:tool-use、api-calling、self-supervision
  • 标题中文:Toolformer:语言模型自学使用工具
  • TLDR中文:本文提出 Toolformer,训练其决定调用哪些 API、何时调用、传入什么参数,以及如何将结果最佳地融入后续 token 预测,在多种下游任务上显著提升零样本性能。
  • 来源文件
  • [OpenAlex discover]
  • [S2 enrich]