CodeScope: An Execution-based Multilingual Multitask Multidimensional Benchmark for Evaluating LLMs on Code Understanding and Generation.
Weixiang YanHaitian LiuYunkun WangYunzhe LiQian ChenWen WangTingyu LinWeishan ZhaoLi ZhuHari SundaramShuiguang DengPublished in: ACL (1) (2024)
Keyphrases
- multi task
- multitask learning
- multi task learning
- learning tasks
- code generation
- text generation
- multi class
- multiple tasks
- source code
- learning problems
- multiclass support vector machines
- supervised learning
- real world
- transfer learning
- information gain
- multi view
- learning environment
- bayesian networks
- feature selection
- data sets