Search
ByteDance Unveils Open-Source Multi-SWE-bench to Accelerate Intelligent Evolution of Large Model Code
2025-04-10 /
Read about 0 minute
Author：小编
ByteDance's Doubao Large Model Team has introduced Multi-SWE-bench, the industry's pioneering open-source, multilingual code repair benchmark dataset. Designed to assess and bolster the capability of large models in autonomously rectifying code errors, this comprehensive dataset encompasses seven prevalent programming languages beyond Python. It comprises 1632 real-world repair tasks meticulously curated from GitHub, each undergoing stringent screening and validation processes. This initiative is poised to propel the advancement of intelligent large model code towards a more pragmatic and engineering-oriented trajectory.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
