method · first seen 14 Oct 2026
evaluation-driven benchmark evolution
Continually generating new benchmark tasks from diagnostic feedback on model failures.
Mentions per week
PapersCodeDiscussion
Earliest sources
- EvoProbe: Evaluation-Driven Benchmark Evolution for LLM AgentsPapers · 14 Oct 2026 · Qianhong Guo, Wei Xie, Xiaofang Cai, Shuoyoucheng Ma, Chengxin Hu, Baosheng Wang
Latest sources
- EvoProbe: Evaluation-Driven Benchmark Evolution for LLM AgentsPapers · 14 Oct 2026 · Qianhong Guo, Wei Xie, Xiaofang Cai, Shuoyoucheng Ma, Chengxin Hu, Baosheng Wang