Robust Speech Recognition via Large-Scale Weak Supervision

  • 类型:arxiv
  • 标识:2212.04356
  • 链接:https://arxiv.org/abs/2212.04356
  • 主题:llm-infra
  • 主分类:multimodal
  • 形态:method
  • 被引:8184
  • 被引来源:Semantic Scholar
  • S2被引:8184
  • OpenAlex被引:1179
  • 影响力被引:1131
  • TLDR:When scaled to 680,000 hours of multilingual and multitask supervision, the resulting models generalize well to standard benchmarks and are often competitive with prior fully supervised results but in a zero-shot transfer setting without the need for any fine-tuning.
  • OpenAlex ID:W4311000453
  • OpenAlex DOI:10.48550/arxiv.2212.04356
  • DOI:10.48550/arxiv.2212.04356
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://arxiv.org/pdf/2212.04356
  • OpenAlex更新:2026-08-24
  • 待LLM分类:否
  • 标题中文:Robust Speech Recognition via Large-Scale Weak Supervision
  • TLDR中文:当将监督规模扩展到 680,000 小时的多语言、多任务数据时,所得到的模型在标准 benchmark 上泛化良好,在 zero-shot transfer 设置下常可与此前全监督方法的结果相当,且无需任何微调。
  • 来源文件
  • [OpenAlex discover]
  • [S2 enrich]