1 paper
Zetong Xiong, Qiao Zhao, Jun Zhang +20
Coding-agent benchmarks increasingly cover long-horizon, end-to-end, and interactive development, but typically retain one requested outcome or a fixed change sequence. Sequential…