1 paper · 1 filter
Jiasheng Zheng, Xin Zheng, Boxi Cao +8
Code sandboxes have emerged as a critical infrastructure for advancing the coding capabilities of large language models, providing verifiable feedback for both RL training and eval…