Recent Papers / arXiv:2606.01286
BenchEvolver: Frontier Task Synthesis via Solution-Centric Evolution
Yangzhen Wu, Aaron J. Li, Wenjie Ma, Li Cao, Ziheng Zhou, Mert Cemri, Shu Liu, Yuran Xiu, Chenxiao Yan, Haikun Zhao et al.
Abstract
Evolves saturated coding benchmarks into harder variants; LiveCodeBench-Plus restores discrimination (Pass@1 27.5–62.6% for frontier models).
RL on evolved tasks improves held-out coding by +8.7 points.
Tasks
editResults
No benchmark results recorded yet.
Benchmark results referencing this paper haven't been added to the registry yet. If you have a reproduction, submit it →
CodeSOTA extraction
Benchmark evidence
- BenchEvolver: LiveCodeBench-Plus Pass@1 for frontier models (extract from Table 2).