Unverified50% confidenceFactExact time
Big Bench was originally a large-scale language model evaluation framework initiated by Google, covering hundreds of tasks to measure models' comprehensive capabilities including reasoning and language understanding.
1
Sources
50%
Confidence
Long-term
Relevance
9/9/2026
First Seen
Sources
Related Entities
Related Claims
Unverified大语言模型随参数规模增长突然涌现出思维链推理能力，由Google的涌现能力论文系统记录75% similarUnverifiedBIG-Bench 是由 Google 主导、超过 450 位研究者协作构建的基准测试集，包含 200 多个任务73% similarUnverifiedGoogle的Gemini采用SentencePiece框架配合Unigram语言模型算法进行分词68% similarUnverified工具增强语言模型（Tool-Augmented Language Model，TALM）最早的系统性论述可追溯至 2022 年 Google 发表的同名论文64% similarUnverifiedGoogle Dialogflow使用了BERT等预训练模型进行意图分类和槽位填充63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/883442API
curl https://kongchang.com/api/v1/knowledge/claims/883442MCP
get_claim(id=883442)